Research on IPTV... and surroundings

Networking: Four ways to reinvent the Internet : Nature News

2010-02-05T11:01:00.000Z

The Internet is struggling to keep up with the ever-increasing demands placed on it. Katharine Gammon looks at ways to fix it.

In Nature

Manifesto for government data

2010-01-25T17:12:00.003Z

The inventor of the WWW (Tim Berners-Lee) and Professor Nigel Shadbolt explain why they decided to create data.gov.uk, here in The Guardian.

Some nice reads from Comms magazine...

2010-01-19T16:32:00.005Z

...on the history of the simplest medium access protocol, Aloha: AlohaNet - Surfing for Wireless Data;

...on the future of IPTV;

...and two nice surveys, one on opportunistic routing and the other on wireless body area networks.

3DTV is coming soon

2010-01-07T13:26:00.002Z

In a surprising endorsement for 3D display technology,Sony Corp. of America, Discovery Communications andIMAX Corp. have announced plans to form a U.S. televisionnetwork entirely devoted to 3D programming.
The three parties have signed a letter of intent to form the unnamed venture, which is scheduled to launch in 2012.
USA Today also reported Monday that ESPN plans its own 3D network, potentially testing the waters for an ABC/Disney 3D network of its own down the line. ESPN 3D will show 80 or so events in 2010, including the World Cup and some NBA and college basketball and football, the paper reported.
in PCMAG.COM

History of OFDM

2009-12-15T23:12:00.002Z

A nice article in the November issue of IEEE Communications Magazine on the history of OFDM:

CoNEXT 1 - SIGCOMM 0

2009-12-05T14:43:00.004Z

Having participated in both conferences, I have to say that imho CoNEXT won this year. Contrary to SIGCOMM, I found most papers very interesting (I would say that 3 out of every 4 papers were very nice). I think it is fair to say that CoNEXT can definitely be considered one of the leading conferences on data communications and networking, side by side with SIGCOMM.

CoNEXT, Day 3

2009-12-03T17:07:00.002Z

Again, some cool and useful papers, and a nice panel in the end of the day moderated by Paul Mockapetris (the inventor of DNS). Other info: next CoNEXT will take place in Philadelphia, from Nov. 30 to Dec. 3, 2010.

Content Availability in Swarming Systems: Models, Measurements and Bundling Implications

D. Menasche (University of Massachusetts at Amherst), A. Rocha (Universidade Federal do Rio de Janeiro), B. Li (Tsinghua University), D. Towsley, A. Venkataramani (University of Massachusetts at Amherst)

This paper received the best paper award.

Bittorrent: huge success because of the self-scalability property. However, there is a problem: availability of less popular content.

Swarm: set of peers interested in the same content. Publisher (seed): interested in dissemination. Peers (leechers): interested in download.

Their setup: monitors in around 100 planet nodes. 45 thousand swarms monitored for 8 months.

Finding 1: content unavailability is severe; 40% of swarms unavailable 50% of the time. Swarms are more available during the first month after they are published.

Finding 2: bundling is common (a single swarm consisting of multiple files – songs in an album, episodes of a TV shown). In music around 75% are bundles.

Finding 3: bundled content is more available. Among all books for instance 62% swarms had no seeds. But in collections of books the figure drops to 36%. “Friends” episodes: 52 swarms. 23 with at least one seed (from these 21 bundles), and 29 with no seeds (only 7 bundles).

For the availability model, they ask the question: “How long is the content available without a publisher?” They use an M/G/inf queuing model for this.

Other questions the model tries to answer: What is the fraction of time in which content is unavailable? What is the mean download time of peers?

Questions for the validation part: 1) can bundling reduce download time? 2) What is the impact of bundling if different files have different popularities?

Answers: 1) bundling reduces download time (idle waiting plus active download time) but only to a certain instant. The model captures this trend. 2) Bundling reduces download time for unpopular time at the expense of popular content.

FairTorrent: Bringing Fairness to Peer-to-Peer Systems

A. Sherman, J. Nieh, C. Stein (Columbia University)

Problem: 1) free riders and low contributing peers consume bandwidth and cause slower downloads, and 2) high contributors lose. Not fair.

Inherent flaw of existing schemes: all based in rate allocation. Inherently imprecise, perform poorly in realistic scenarios. What if we don’t use rate allocation?

Idea: FairTorrent algorithm. 1) For leechers: each leecher maintains deficit counter with each other leech. Always send data to leecher with lowest deficit. The effect is that it ensures fast rate convergence. 2) For seeds: pick next leecher using round roubin. Evenly splits bandwidth among leechers.

Properties: 1) fast rate convergence of upload/download rates, 2) resilience, 3) fully distributed, 4) requires no estimates of peers intended rates allocation.

Results: Really fast convergence compared to other systems. Download times also better.

More info in http://www.cs.columbia.edu/~asherman/fairtorrent

Scaling All-Pairs Overlay Routing

D. Sontag, Y. Zhang (MIT), A. Phanishayee, D. Andersen (Carnegie Mellon University), D. Karger (MIT)

The goal: find best routes in full-mesh overlay networks. Motivation: rapidly route around fails (resilient overlay networks) or find lower latency (skype).

Prior approaches to scaling routing: choose from a few alternate paths (by eliminating the high latency ones for instance). They show how to find optimal routes in a more scalable manner.

Scaling to large overlay networks. Where is the bottleneck? 1) each nodes probes all other, 2) per-node communication thus scales linearly. Link state routing sends info to all nodes. Quadratic scaling.

They present an algorithm that finds all optimal routes with only nsqrt(n) per-node configuration. They prove in the paper that their algorithm is theoretically the best possible.

T heir approach to one hop routes: send a link state to a subset of nodes only, and not all. These are the rendezvous nodes (RN). Then each RN will calculate its best interface to all places he received info for, and then send this recommendation for the same guys that sent stuff to them. So, with fewer rendezvous servers, you have less communication.

How to assign rendezvous servers? 1) every two nodes should have a rendezvous server in common. 2) They use quorum systems to provide this property (they use grid quorum). Each node has 2sqrt(n) rendezvous servers.

Summary: round 1) size n link state tables sent to 2sqrt(n) nodes; round 2) size 2sqrt(n) recommendation tables to 2sqrt(n) nodes.

Strategies to deal with RS fails: built in redundancy (every pair of nodes has 2 RS in common).

Multi-hop algorithm: repeat the one-hop algorithm using modified link costs.

They applied the algorithm to find optimal one hop routes in a Resilient Overlay Network (RON). Key questions: how do rendezvous failures affect bandwidth requirements? They deployed this in planetlab with 140 nodes for 2 hours. They observed link failures: on average 50% of nodes experience over 8 concurrent link failures. Double rendezvous failures are infrequent.

StarClique: Providing Privacy Guarantees in Social Networks using Graph Augmentation

K. Puttaswamy, A. Sala, B. Zhao (University of California, Santa Barbara)

Social apps are growing. You share valuable info with friends (recommendations, social bookmarking). Example: social web caching app. By sharing web cache with friends, you improve performance and also it works as a collaborative phishing filter.

Major concern of the user using these apps: privacy. 1) Friends receive data, but if a friend’s computer is compromised, then the info can be used for other purposes; 2) botnets are targeting social network credentials; 3) effective spamming via social engineering; 4) rising of effective phishing attacks.

Naive defense: remove user identity from the data shared. But it is still possible to have a social intersection attack: several attackers intercept the social network topology. Properties of the attack: 1) need 2 or more attackers, 2) application independent, 3) works within the applications bounds.

They measured the attack’s effectiveness (in facebook), and realised that more than 70% of the users can be exactly identified by other users... so the attack can be really effective.

Now, solution to defend to this attack. 1) Intuition: increase the number of common friends (more common friends lead to better privacy); 2) adapt the k-anonymity approach; 3) anonymise the social graph to achieve k-anonymity

Cost of solution: users receive additional data, and app servers process and send additional data

Goals of their system model: minimise number of latent edges added, and keep the latent edges close in social distance. They evaluated their system in 6 different traces from previous studies (facebook, flickr, etc.).

Lockr: Better Privacy for Social Networks

A. Tootoonchian (University of Toronto), S. Saroiu (Microsoft Research), Y. Ganjali (University of Toronto), A. Wolman (Microsoft Research)

The state of privacy on Online Social Networks (OSN): 1) no control over what facebook does with my social info; 2) users need to duplicate their social info; 3) must register to see friend’s content; 4) privacy terms can be contradictory, confusing, and make some users very nervous. Sense of privacy is really not very good.

Lockr’s goal: today users have no control of their social info. They want to put the user in control of their info. How? By using social attestation. You can say that a certain user is your “friend”, or “colleague”, or “drinking buddy”. The social network is created by passing out attestations. You can store your content anywhere, but you set a social Access Control List when you upload this stuff. Everything revolves around you. Also, there is an attestation verification protocol (for you to download info from a user).

Locker works with any OSN. Also works with P2P systems, for instance.

Privacy benefits: 1) users form one social network using social attestation; 2) they continue sharing content via the normal OSN or P2P mechanisms.

What privacy problems are solved by Lockr? 1) “users have no control of their social info”: with lockr users can choose where to store their attestations; 2) “user must register to see friend’s content”: in lockr it removes this, users just show their attestation(only requirement is to map ID to public key); 3) “social deadlock” (in P2P each user wants to verify the other attestation’s first), or “who goes first?”: in lockr you also have a relationship key (all your friend’s have your “friends” relationship key); 4) “hard to revoke attestations”: in lockers the attestation has expiration dates; 4.1) “expired attestation still have valid relationship keys”: a) rotate relationship keys daily (until it expires) – but, lots of keys, so 2) they use a one way hash chains; 5) “OSN sites can now ‘sell’ attestations”: they use zero knowledge proofs for attestation verification. So you keep a signature of your attestation (not sharing the signature). But you can prove to anyone that you can sign the attestation, without sending the signature.

Is lockr a real system? Yep - they implemented Lockr center as a facebook app, and added lockr for bittorrent and flickr.

Code: http://www.lockr.org

ARES: An Anti-jamming REinforcement System for 802.11 Networks

K. Pelechrinis, I. Broustis, S. Krishnamurthy (University of California, Riverside), C. Gkantsidis (Microsoft Research, Cambridge)

Launch DoS attacks in WiFi nets in easy. Jammers can disrupt communication. How to deal with jammers? Frequency hopping was a possibility, but this is weak (WiOpt2009: only 4 jammers can block entire 5GHz spectrum).

So, how to alleviate this without frequency hopping?

Main contributions: 1) fixed rates often preferable in the presence of a jammers; 2) clear channel assessment (CCA) is important.

1) Interaction random jamming/rate control. Jamming effects last beyond the jamming period when rate control is used. In the presence of an intermittent jammer the rate control algorithm might be slow to converge to optimal rate. Remedy: fixed rate assignments increase immediately throughput. So it is important to decide when to perform rate adaptation. This can be done analytically. They show that fixed rates provide high throughput gains. However, perfect knowledge is not realistic, so they include a Markovian Rate Control (MRC) without requiring knowledge of any of the parameters needed for than analytical solution. This can be tuned to be close to optimal.

2) Rate control removes transient jamming effects. What about constant jamming effects? Solution: power control. Power adaptation helps only when transmitter is not in the jammer’s range, and low transmission rates are used. Increasing power helps when transmission rate is low.

How to deal with high power jammers (compared to a legitimate wireless card)? Transmitter needs to ignore jamming signals, and receiver must decode legitimate packet. Using CCA helps in here. There are, however, side effects to this: the transmitter may become a jammer unintentionally, and the receiver can look at legitimate signals and think they are noise. So, they present a simple heuristic for setting the CCA on the link.

By using these observations on rate and power control they built ARES. It works like this: if jammer is detected, check if CCA is tuneable. If so, they invoke power control. If it doesn’t solve the problem, do rate control. If it is not tuneable, just do rate control.

They implemented their system in an indoor testbed. With rate control only it improves performance up to 100%. With a mobile jammer the system can increase performance (using power control) by 150%. Note: by adding more APs the benefits are reduced.

ThunderDome: Discovering Upload Constraints Using Decentralized Bandwidth Tournaments

J. Douceur, J. Mickens, T. Moscibroda (Microsoft Research), D. Panigrahi (MIT)

Motivation: upload constraints limit scale performance in online games, for instance. How to discover these constraints? Bandwidth estimation is non trivial...

ThunderDome provides estimates that are fast, accurate and decentralised.

Many tools for one way bandwidth discovery: packet rate (pathload, PTR). What is the problem with these? 1) Directional ambiguity; 2) what’s the proper probing schedule?, 3) how to deal with probing error?

Solution: 1) pair-wise bandwidth probe: we measure b(X->Y and Y->X) between pairs of nodes to determine upload speed; 2) decentralised bandwidth tournaments, 3) probe aggregation.

Participant model assumptions: 1) peers gather and then stay for 10s of minutes: ex. games. 2) hosts have public IP addresses or cone NATs, 3) hosts do not lie. Network model assumptions: 1) peers connected by hub and spoke topology, 2) asymmetric bandwidth constraints.

The pair-wise bandwidth probe: hosts X and Y transmit to each other. With a single exchange they can get the upload speed of the loser. You do these between pairs of nodes (the “tournament”), and in the end you have info of all (except one).

But we may have measurement errors: they may misjudge one way bandwidth, confusing the upload of two hosts. This occurs when separation between up and down speed is too small (quite common in dial up). To recover from the problem we lose nodes in the “tournament”: basically, we add more samples.

They also have a system to coordinate the “tournament”, basically building a tree. These coordinators can be regular peers, so it is mostly decentralised.

Where the Sidewalk Ends: Extending the Internet AS Graph Using Traceroutes From P2P Users

K. Chen, D. Choffnes, R. Potharaju, Y. Chen, F. Bustamante (Northwestern University), D. Pei (AT&T Labs Research), Y. Zhao (Northwestern University)

Currently the public available AS graph is incomplete. ISPs do not want to disclose their connectivity.

Existing approaches: passive (BGP routing tables/updates), active (traceroutes). Limitation: small number of vendor points (ASes with monitors inside).

Promising platform should: 1) have more VPs, 2) allow active measurement, 3) scale with the Internet growth. So, P2P seems a natural fit.

Their approach: 1) increase the number of VPs using P2P users, 2) active traceroutes from edge of the internet, 3) map IP paths to AS paths, extract AS links. They also offer many corrections to false AS links, bad traceroute results, etc.

It seems a more complete AS map is now available online, which is cool:

http://aqualab.cs.northwestern.edu/projects/SidewalkEnds.html

Exploiting Dynamicity in Graph-based Traffic Analysis: Techniques and Applications

M. Iliofotou, M. Faloutsos (University of California, Riverside), M. Mitzenmacher (Harvard University)

Focus on the network wide behaviour of internet apps (IP to IP communication patterns), and build Traffic Dispersion Graphs (TDs). Goal of these graphs is to do behavioural profiling (assign apps to different groups, and detect unusual behaviours).

Previous work: static graphs to capture network-wide behaviour. These can distinguish client-server from p2p apps. But how to distinguish different flavours of these two?

In here they use dynamic graphs characteristics to extract more app properties. It is dynamic because nodes and edges change with time. Questions that arise: how to express the dynamicity of traffic graphs? What metrics to use?

How to express dynamicity: 1) graphic represented as a sequence of graph snapshots taken over time, 2) use metrics that quantify differences between graphs.

Metrics: static, like average degree, degree distribution, etc., but also binary ones that allow checking how nodes or edges change with time.

To quantify dynamic behaviour, they create a new metric: 1) detailed edge similarity, which shows both persistent and non persistent edges, for instance.

Application: detecting polymorphic blending. Action: an app mimics the traffic of another app to avoid detection. Effect: traffic of polluting app will be added to the target app. This will change the dynamics of the graphs.

They looked at many metrics (both static and dynamic), to see which detect polymorphic blending better. For small blending the dynamic metrics are more sensitive and result in more alarms. However, the best is to mix the static and dynamic metrics.

RFDump: An Architecture for Monitoring the Wireless Ether

K. Lakshminarayanan, S. Sapra, S. Seshan, P. Steenkiste (Carnegie Mellon University)

Popularity causes crowding. Look at WiFi unlicensed spectrum: what if we turn the microwave oven? How to troubleshoot these problems?

Sniffers don’t work well in the physical layer due to this interference of all sources. How could we sniff such wireless network? A possibility would be to add lots of interface cards (WiFi, Bluetooth, etc).

What about using software defined radio (SDR)? SDR hardware exposes physical layer info and supports programmable analysis modules.

Challenges of SDR: 1) how to process 256 Mbps of info in real time? 2) How to differentiate between samples? The requirements: 1) real time, 2) multiprotocol, extensibility.

Other naive solution: demodulate all signals received. Good thing is that it is protocol extensible. Problem: real time capability missing. Demodulation is costly, so all demodulators have to process everything. How to make it more efficient?

A better solution: use a energy filter after the SDR. This filter decides what signals it can use (this can remove noise, for instance). Demodulators do less work. But it is only OK when medium utilisation is low. If it is high, it can’t work on not real time.

Their solution: RDump. After the energy filter, insert a fast detector. This detector detects to which demodulator the signal should be sent. This solves everything: it is protocol extensible, and real time. Detectors can be faster because they can tolerate false positives, and can tolerate delay.

The tricky thing is how the detector works. How to detect which protocol it is? 1) Check time: 802.11 uses SIFS and DIFS, Bluetooth uses TDD slots; 2) check how phase changes: 802.11b uses DBPSK, Bluetooth GMSK; 3) also check frequency: look at channel width, 802.11b uses 22MHz, Bluetooth 1MHz.

They implement this in GNU radio and USRP SDR platform, and achieve pretty nice results.

Panel: What will be the origin of the next network revolution?

Moderator: Paul Mockapetris (inventor of DNS, ISI)

Members: Bruce Davie (Cisco), Van Jacobson (one of the main contributors to TCP/IP, PARC), Dirk Trossen (BT Research), Kenjiro Cho (IJJ Japan), Max Ott (NICTA)

More info here: http://conferences.sigcomm.org/co-next/2009/panel.php

In general, everyone seemed to agree that research needs more money, that betting in risky projects is important (not only incremental research), that more cooperation of industry is needed (it is increasingly going down), and that more freedom is needed (don’t expect great results in six months).

An interesting point by Keshav: it is easy to get money from industry, but then the company doesn’t have anyone to follow the research project. “No one answers my calls!” The company seems to give the money but then doesn’t care with the results. Bruce Davie agreed, saying that it is cheaper for Cisco to give that money “away” than to find a person inside the company to follow the research work.

CoNEXT, Day 2

2009-12-02T19:59:00.000Z

Another very interesting day at CoNEXT. Van Jacobson presented his extremely interesting CCN work (definitely the highlight of the conference – maybe this will be known as the “CCN conference” in the future... :). Besides that, other very interesting talks too.

Keynote Presentation by the 2009 ACM Sigcomm Rising Star Ratul Mahajan (Microsoft Research, USA)

The title of the talk was “How to build research network systems in your spare time”.

The goal: articulate a research method for building network systems.

Method:

0. Pick domain (be wary of the hot trends);

1. Know the problem well (papers alone are rarely a good source – scrutinise, measure, survey);

2. Debate several solution ideas (read more papers for solutions – from other areas for instance – than for problems);

3. Start small and then embellish (to evaluate, see if it works, when and why it works, and what are the benefits and costs);

4. Make it real (release data and code, deploy, talk to practitioners).

Networking Named Content

V. Jacobson, D. Smetters, J. Thornton, M. Plass, N. Briggs, R. Braynard (Palo Alto Research Center)

For 150 years, communication has meant a wire connecting two devices. For users, the web forever changed that; content matters, not the hosts it came from.

Design criteria 1: IP runs over anything (that moves bits to some destination), anything runs over IP. CCN runs over anything (that moves bits in space and time), anything runs over CCN. But you don’t need any entity in layer two. You just talk. There is no state to establish. Notion of communication generalises.

Design criteria 2: an evolutionary path: a) incremental deployment gives immediate value, b) painless, viral, bottom up growth, c) use existing infrastructure, including its lessons.

Design criteria 3: security (communication = trusted info). Trust has to be a property of the info, not the path it comes over or the container it’s in. Any entity should be able to assess the integrity, relevance and provenance of any piece of info.

Basic idea: 1) consumer announces an interest over any and all available communications links (an http get kind of thing). 2) Interest identifies a collection of data – all data items whose name has the interest as a prefix. 3) Anything that hears the interest and has an element of the data sends it.

Answer to some “Buts”: “This doesn’t handle conversations or realtime” – yes it does – see Rearch voccn paper yesterday.

“this is just google” – this is IP for content. They don’t search for data, they route it.

“this will never scale” – hierarchically structured names give same log(n) scaling as IP but CCN tables can be much smaller since multi-source model allows inexact state (e.g., bloom filter) – tradeoff: a little bit of bandwidth can save state.

Measurements: on an “apples for apples” comparison, rise time is worse than normal TCP, and achiever lower throughput because there are more headers for signatures (but not that bad). But, on shared content distribution, that’s where CCN wins very clearly.

Other cool stuff: 1) names can be context sensitive and active. 2) Robust, simple single point configuration (you are not talking to something). 3) Security model makes infrastructure easy to protect and hard to attack. 4) Unifies communication and storage (files and wires are really the same thing).

Open source, GPL’d, CCN stack and VoCCN linphone at www.ccnx.org.

A comment from the audience resumes the expectations this work is generating: “It has excited me more than anything I have heard in the past 10 years”.

Virtually Eliminating Router Bugs

E. Keller, M. Yu (Princeton University), M. Caesar (University of Illinois at Urbana-Champaign), J. Rexford (Princeton University)

Routers have complex software, so they have bugs (XORP has 826k lines... CISCO routers have millions of lines).

Main question: how to detect and stop the problem caused by the bug before it spreads?

Answer: Run multiple, diverse routing instances (software and data diversity ensures correctness - SDD).

SDD challenges in routers: 1) making replication transparent to neighbour routers (they can be normal routers), 2) handling transient real time nature of routers (react quickly to network events, but do not over react).

SDD opportunities: 1) easy to vote on standardised output, 2) easy to recover from errors via bootstraps (routers depend very little on history), 3) diversity is effective in avoiding router bugs.

Hence, outline of the talk: 1) exploiting SDD, 2) bug tolerant router architecture, 3) prototype and evaluation.

Why does diversity works? 1) There is enough diversity in routers (software: quagga, xorp, etc.), protocols (ospf, is-is), 2) enough resources for diversity (extra processor blades for hardware reliability, multi-core processors), 3) effective in avoiding bugs

Evaluate diversity effect: most bugs can be avoided by diversity. 88% of the bugs avoided in their experiment.

Effect of software diversity: by picking 10 bugs from xorp and quagga they realised that none were present in the other implementation. Also, static code analysis on version diversity: some bugs from one version are corrected, others added, but in the end they don’t overlap that much.

Bug tolerant router: protocol routing instances running in parallel. Then when an update arrives, there is a voter that votes on the results, generating a single setup to the forwarding table. So it is transparent model (it intercepts system calls only).

How to do the voting (two shown, others in paper): 1) master slave approach (speed reaction time, but doesn’t handle transient behaviour well) and 2) continuous majority (handles transience better).

Voting code is small (500 lines of code)

Prototype and evaluation: 1) no modification of router software, simple hypervisor, built on linux with xorp and quagga. 2) Evaluated using BGP trace from Router Views. 3) Evaluation checking delay introduced.

Test with 3 XORP and 3 quagga routing instances. Single router: 0.066% fault rate. Master slave (reduces to 0.0006%, with delay of 0.02 seconds) and with continuous majority reduces even more, with a bit higher delay.

Small overhead: Delay overhead of hypervisor: 0. 1% (0.06sec). Overhead of 5 routing instances (4.6%).

More info on http://verb.cs.princeton.edu/.

MDCube: A High Performance Network Structure for Modular Data Center Interconnection

H. Wu, G. Lu, D. Li, C. Guo, Y. Zhang (Microsoft Research Asia)

Mega data centre: millions of servers. Each container has thousands of servers, and they want to connect hundreds of these.

Design goals: 1) high bandwidth for data intensive computing, 2) cost effective, 3) addressable cabling complexity

Existing approaches: 1) tree (oversubscribed, bottlenecked root), 2) fat tree (multiple rooted tree, costly as large number of switches)

MDCube design: leverage the link rate hierarchy of Ethernet switches (1GBps for server switch port, and 10gbps to 40gbps for switch uplink). Use switch uplinks for intercontainer interconnection.

Construction of MDCube: 1) treat each container as a “virtual node”, 2) treat switches high speed uplink interfaces as “virtual interfaces”, 3) connecting switches to peer switches directly, 4) virtual nodes form a modular data centre cube: MDCube. Inner container structure: BCube (presented at sigcomm).

With 48x1G+4x10G they can support 5M+ servers...

Properties of MDCube: 1) small network diameter,2) high capacity between servers, 3) large ABT (aggregate bottleneck throughput), 4) graceful incremental deployment.

MDCube routing: 1) hierarchical routing (divide and conquer), topology is known; 2) loosely controlled source routing with load balancing and fault tolerance. Detour routing: similar to Valiant Load Balance. In VLB, a random node is selected. But they choose to detour only at container level. Fault tolerant routing: hierarchical routing approach (source node only chose which container to route, and path inside container is maintained by BCube).

Greening the Internet with Nano Data Centers

V. Valancius (Georgia Institute of Technology), N. Laoutaris (Telefonica Research), L. Massoulie, C. Diot (Thomson), P. Rodriguez (Telefonica Research)

Data centres at the moment host applications. Problems: expensive to build, expensive to run. Far from user. Lots of redundancy for robustness. Power usage: use lots of power, for example for cooling, and underused resources.

Nano datacenters (NaDa): some content servers (not as many as today), and a nano data centre network (p2p), with gateways. Benefits: ISP friendly (reduce load on backbone), self scalable (adding a new user also joins in a guy to supply content), saves energy (reusing baseline power, no need for cooling and redundancy, reduce energy consumed in network).

Challenges: limited uplink bandwidth, scale, incentives for users, data privacy, cost of gateway.

Focus of this talk on how to save energy.

Proportional Energy use in NaDa: currently servers have a not very nice power usage profile (big baseline power). With NaDa, using dedicated home gateways for NaDa service, would be even less efficient then servers. But we can amortise the baseline energy using gateways only when they are active for something else. So power consumption is proportional to load.

Evaluation: 1) energy measurements of content streaming servers, residential gateways, and DSLAMs. 2) Content access traces (from commercial IPTV provider, youtube, netflix), 3) gateway availability (lower and upper bounds from measurements).

Event driven simulation, with various parameters: 1) upstream bw, 2) storage, 3) video streaming rate, 4) number of NaDa users, 5) network distance, 6) data centre cooling and energy distribution losses, 7) residential energy distribution losses.

Measurements: server vs gateway. Carrier grade IPTV streaming server – server energy use 211 joules /gbit; 2) Thomson residential DSLM gateways: DSLAM power use unaffected, gateway energy use: 100 joules /gbit.

Most important variable for NaDa: bandwidth. Uptstream rate (UR) and video rate (VR). When UR>VR everything fine: each gateway can serve more than one user, and increased replicas used. If upstream bandwidth is increased we can get more energy efficiency in networks.

SWARM: The Power of Structure in Community Wireless Mesh Networks

S. Das (Qualcomm), K. Papagiannaki (Intel Labs), S. Banerjee (University of Wisconsin at Madison), Y. Tay (National University of Singapore)

The work is aimed at community wireless networks (meraki, roofnet, CUWIN), where we have wireless forwarding nodes (mesh routers) and few gateway nodes with network connectivity. The clients connect to a mesh router for access to a gateway.

Observations: 1) mostly single radio networks (low cost, off the shelf); 2) traffic patterns mainly between gateways and mesh routers; 3) this is a static/stable network and the objective is to find gateways to access internet, so using normal ad-hoc techniques is not good.

Problems with current approaches: 1) do not optimise network topology for the intended use of the network; 2) interference not taken into account; 3) requires operation on same channel.

Motivations for SWARM: 1) improve performance by organising the network better; 2) gateway trying to minimise interference; 3) nodes assigned to gateways for good performance given a certain network scenario.

Their scheme involves 3 phases. Phase 1: enumeration (iterate through all possible assignments of MRs to gateways) and pruning (for disconnected or imbalanced configurations). Phase 2: Construct a tree (simple interference aware model). Phase 3: channel assignment (extension of MAX-chop algorithm for WLANs, mobicom 2006): choose channel hoping sequence that minimises conflicts.

Evaluation by simulation in qualnet and also a deployment on a testbed shows that SWARM increases throughput significantly.

EZ-Flow: Removing Turbulence in IEEE 802.11 Wireless Mesh Networks without Message Passing

A. Aziz (EPFL), D. Starobinski (Boston University), P. Thiran, A. Fawal (EPFL)

Multi-hop scenario: greedy sources, with no source rate limiting. The intermediate nodes are relays, and a single channel with CSMA/CA, and with turbulence (instability – this problem occurs in a stable wireless network larger than 3 nodes).

Design goal: network stabilisation (delay reduction), unmodified MAC layer, backward compatibility.

EZ flow: analogy to road traffic problem (same as a solution in New York, to generate a “flow of green lights”: EZ-pass). It is useless to send more packets to a heavily loaded successor.

EZ-flow lies just above the MAC layer. It dedicates one MAC queue per successor (max. 4). It is composed of 2 modules: Buffer Occupancy Estimator (deriving successor buffer) and Channel Access Adaptation (adapts the cw accordingly).

BOE module: works passively without any message passing. This is the main advantage of the scheme. Also, no header modification (hence, runs with current hardware).

How to derive buffer occupancy implicitly: 1) keep in memory a list of identifiers of recently sent packet. 2) use broadcast nature of the medium to monitor the forwarded packet. 3) obtain (precisely) the current buffer occupancy at successor.

Why is explicit messaging less precise? Packet header can be modified just before going to the MAC, and all this adds queuing delay, so the info can be out of date (that doesn’t occur in the passive option).

CAA module: 1) adapt channel access probability my modifying cw. 2) decision based on successor buffer size, derived by BOE, plus some thresholds.

Nice results in reducing delay and in stabilising the network.

The Age of Impatience: Optimal Replication Schemes for Opportunistic Networks

J. Reich (Columbia University), A. Chaintreau (Thomson)

Infrastructure-based content dissemination drawbacks: expensive, not always available, capacity can’t keep up with demand.

Opportunistic content dissemination: each meeting of nodes is a chance to transfer data. Potential: huge pool of untapped bandwidth; doesn’t require infrastructure. But: contacts unpredictable, may be much slower, limited local memory and energy buffer

Problems: 1) how to chose what content should be placed where (in which peer) to satisfy impatient users? 2) How to attain a desired allocation with an opportunistic p2p protocol?

Contributions of this paper: 1) they show what is the optimal cache allocation for impatient users (analytic model), and 2) how to apply these insights in a distributed scheme?

Impatience: increasing wait, decrease utility. Their objective function is the aggregate expected delay utility (U). They choose an allocation (X) to maximise the objective – this minimises losses due to impatience.

Main analytic results: 1) objective U is submodular in X. 2) In a homogenous network U is concave.

Their distributed system QCR is demand/cache reactive: depending on the impatience factor (can the user wait for, say, one hour for the service?) they defines parameters like cache size.

Feasibility of content dissemination between devices in moving vehicles

T. Zahn, G. O'Shea, A. Rowstron (Microsoft Research Cambridge)

Can we use opportunistic WiFi communication to disseminate content between PND (Personal Navigation Devide) in vehicles (assuming no static infrastructure, AP or cellular)? Most prior work involved infrastructure.

Challenges for vehicle to vehicle communication: 1) initial density low (thus exploit every encounter), 2) short connections (e.g. passing side-street), 3) variable link quality.

Why PNDs? 1) Lots of them around, good storage capacity; 2) not just road maps (also gas stations or restaurants info, other services, etc.).

How to update data? 1) USB cables; 2) built-in GSM/3G (still expensive); 3) bluetooth to cell phone (suitable contract and data plan needed – some cost); 4) FM radio – low bit rate. So: what about WiFi?

Design: 1) assume a global set of versioned files; 2) when 2 devices meet a) rapid device detection over ad-hoc network, b) find a common file with different versions, c) send data using light weight FTP.

Phase 1: rapid peer detection – join or form 802.11 ad-hoc network (fixed channel).

Phase 2: discovery protocol: use of bloom filters to a) hash some of its filenames, and also b) hash filename and versions. Then send this message to neighbour. Receiver does the same, and sends back similar message. Then make the transfers of the new versions of files between them.

Phase 3: transport protocol. Block oriented transfer of files of fixed max size. Ack is a bit vector.

Evaluation with their raw protocol and TCP/IP. They can transmit significant amounts of data even for vehicles moving at considerable speed.

Cool-Tether: Energy Efficient On-the-fly WiFi Hot-spots using Mobile Phones

A. Sharma (University of California, Santa Barbara), V. Navda, R. Ramjee, V. Padmanabhan (Microsoft Research India), E. Belding (University of California, Santa Barbara)

There are more mobile users than broadband users: so it is probably better to provide internet connectivity using mobile phones.

Motivation: 1) provide ubiquitous connectivity to devices, 2) leverage smartphones as internet gateways (considering battery life), 3) multiple phones around you at home, work, or on the move.

Other tethering mechanisms: 1) via usb (not possible to use multiple phones), 2) WiFi adhoc mode (no power save mode), 3) via Bluetooth (short range, low data rates, hence high energy/bit cost)

Design goals: 1) support multiple gateways, 2) reduce energy consumption (optimise LAN and WAN).

Incremental cost of sending more data is negligible. Hence, sending at highest possible rate is energy efficient. Also, activating an interface cost a lot in terms of energy, so it is probably not good to use many mobile phones at the same time.

Energy aware design: 1) keep WAN pipe full whenever in use. For this they aggregate web requests in a gatherer present on the internet (the cloud deals with all the requests a single web page request needs to make, since one page has many links inside requiring several requests). This way we transform the short bursts into a single high burst), 2) use optimal number of number of phone (they’ve done some analytical work to find the optimal number of mobile phones), 3) use reverse mode WiFi: laptop is the AP, mobile phones the clients.

Based on 1) (keep WAN pipe full) they get very nice savings in energy, and also reduce latency. They compare their results with COMBINE, and they achieve 38% to 70% savings compared with this scheme.

CoNEXT, Day 1

2009-12-01T18:01:00.001Z

I am attending CoNEXT these days, and as usual I will maintain some notes on the talks I find more interesting.

ReArch'09

During the morning I decided to attend the ReArch workshop. Some interesting talks in the morning sessions, with Mark Handley and Van Jacobson figuring at the top of the list.

Keynote Presentation by Mark Handley (University College London)

A similar talk to the one Mark gave when he received the Royal Society-Wolfson Research Merit Award (as far as I can remember).

The Internet Control Architecture can be divided in three parts: routing, congestion control, and traffic engineering. These three things have been a bit disconnected because they were not planned together: they were an accumulation to fixes to specific problems (TCP with Jacobson addition for congestion control, BGP for routing, etc.). The control plane was made of incremental fixes, where the power of legacy had an important role.

It would be nice to have an architecture where all control protocols fit well.

Main question: is it possible to change the internet architecture in a planned way, so as to achieve long term goals?

By looking back to Internet control architecture, are there opportunities to change the game? In routing, it’s nearly impossible to replace BGP. It would have a huge network effect. For congestion control it is nearly impossible to replace TCP. But there is lots of stuff to do in-between these “layers”.

A wish list for control mechanisms: very high robustness (no downtime, robust to attack), load dependent routing (move traffic away from congestion), diverse traffic mix, and sane economics (reward net operators for investment).

Also, consider the problem of multi-homing. Multi-homing provides redundancy (more than one path provides robustness). However, routing is not a good way to access that redundancy. It would be good to access the redundancy at the transport layer, where congestion control can see it. With multi-homing (multipath) servers can do load balancing. Idea of pooling resources: several links working as if they were a single pool. Multipath TCP pools multiple link resources

Economics: what is the marginal cost of a packet? If no one else wanted to send at that time, no cost. It only makes sense to charge if our traffic displaces another user’s traffic. We should be charging for congestion.

Common models – rate based or volume based charging (x giga per month) don’t offer the right incentives. They are inefficient economically. Also, some apps are latency sensitive, some only care about long term throughput. Charging for congestion volume would thus make sense, encouraging machine to machine traffic to move to uncongested times.

ISPs can’t charge for congestion because they can’t see it properly. But end systems can – and some routers too. Bob Briscoe’s re-feedback and re-ECN ideas (congestion exposure): to indicate to an ISP the congestion seen by traffic downstream. Congestion exposure is thus an enabler for sane economics.

VoCCN: Voice Over Content-Centric Networks

Van Jacobson (Palo Alto Research Center (PARC), USA); Diana Smetters (PARC, USA); Nicholas H. Briggs (Palo Alto Research Center (PARC), USA); Michael Plass (Palo Alto Research Center (PARC), USA); Paul Stewart (Palo Alto Research Center (PARC), USA); James D. Thornton (PARC, USA); Rebecca L Braynard (PARC, USA)

Everybody knows that content based networking is great for content dissemination, but can’t handle conversational or real time traffic: everybody is half right... In fact, content networking is more general than IP, and does anything that IP can.

They implemented VoCCN, a VoIP functional equivalent based on CCN to prove this.

VoCCN: why bother? VoIP works badly for multipoint, multi-interface and mobility. VoIP wants to talk to an IP address, and in a voice call we want to talk to a person. Also, VoIP security is poor (only SSL between proxies).

With CCN, no need to map from a user to an address: we have very flexible names. And sender can encrypt its identity, with only the receiver able to decrypt it. No need for the user to reveal its identity. Supports secure VoIP.

They’ve built the voice app on top of CCN (instead of IP), and in the end the app performance was very similar to VoIP. They flipped frantically from one Ethernet connection to another, and the shift was almost imperceptible, and no packets were lost.

All this is available open source: www.ccnx.org. The VoCCN linphone code should be there by the end of the week.

Classifying Network Complexity

Michael H. Behringer (Cisco, France)

Network complexity is increasing: everybody knows that. But what is “network complexity”?

You need some complexity to achieve a certain level of robustness. How to deal with complexity? Divide and conquer (layering, o-o: “classes” matter, not instantiations), shifting complexity (e.g., away from the human – make simple user interfaces – look at the iPhone), meta-languages, structural approaches (reduce dependencies by design).

The “complexity cube” idea: a cube with three axes that represent the three areas of complexity: operator, physical network, network management.

Future work: quantitative metrics, impact of the rate of change, investigate human factors.

IP Version 10.0: A Strawman Design Beyond IPv6

Kenneth Carlberg (SAIC, USA); Saleem Bhatti (University of St Andrews, United Kingdom); Jon Crowcroft (University of Cambridge, United Kingdom)

Once upon a time: internet unknown by the general public, best effort was the only game in town, people used telnet. But now: need for security, and need for a Next Generation IP.

We were running out of address space. CIDR and NAT are near term solutions. Also, associated routing table size explosion. Solutions: new lookup algorithms to reduce impact, and faster hardware. But multi-homing has renewed the problem...

New Generation IP: 1) Simple IP (Steve’s IP) – minimise header, adding more extensibility (one or more Next Header), flow ID, larger flat address structure. 2) The P internet protocol (Paul’s IP) – change addressing to locater and identifier split, and hierarchical and variable length locator (implied source routing). Finally, 3), the grand compromise of 94: the simple IP-Plus. Simple IP with hierarchical addresses of Paul’s IP: IPv6.

Critique to IPv6: not much of an architectural change. 1) Large 128 bit address (retains locator & identifier, provider is still cling to NATs – they have no economic incentive to migrate). 2) Same size diff-serve field. 3) Multiple next headers (encapsulations or MPLS). 4) End to end flow labels (“market” uses islands to cut through routing (MPLS)). Note: recent report show IPv6 traffic is 1/100 of 1% of all IP traffic... Main questions: Does “more” qualify as architectural change? Where are the “must have” features?

Four significant discussions on “location/Identifier split”: 1) 1977 – tcp and mobility, 2) 1992-93: Paul’s IP work, 3) 1996 (O’Dell 8+8 proposal, 4) 2007 (IAB report). Three efforts now: HIP, LISP and ILNP.

Multi-homing problem: provider independent prefixes tend to be popular, but are non aggregatable.

IPv10 design: retain minimalism and extensibility of IPv6. Incorporate identifier/locator split. Besides headers, also introduce tails (change state insertion model: temporary headers and tails...). In the header include header navigation and forwarding info. In the trailer: trailer navigation , end to end info, diff-serv.

Impact of tails: change the end to end model of constructing headers (facilitate temporary insertion of overhead info). Avoid inefficient encapsulation. Forster the need to go beyond current ASIC header lookup limitation.

Should we be more radical in our design? Are there any must have features in IPv10?

ACM CoNEXT Student Workshop

During the afternoon we had the poster session, where I presented the poster “Relative Delay Estimator for Multipath Transport”. After that, a nice panel session to offer advice to PhD students, which included a discussion on what is best: to publish in conferences or journals? Contrary to most other fields, CS researchers tend to prefer conferences... However, there was some important points in favour of journals - namely the idea of creating a "scientific archive". Keshav, for instance, defended strongly journals, and invited everybody to publish their research in CCR... :) Also, TON wants to add bigger size papers and reduce the costs to add extra pages. We may see a move to papers in the meanwhile. A nice idea seems to be to publish in a top conference, and then to publish a longer version as a journal paper. Let's do it! :)

New internet TV announcements

2009-11-15T07:07:00.003Z

An article in IT World mentions several announcements made recently that could make Internet TV a real possibility in the near future.

The first was You Tube, that anounced they will start offering full HD streams (1080p). So far, You Tube has an HD option with a resolution of only 720p. To understand the difference, check this wiki entry.

Next is Boxee. This company will launch a Boxee box that will enable web content to be delivered into your TV. Another possibility is to buy the new Dell Mini Desktop computer and connect it directly to the TV.

Now that you have Internet TV in your, well, TV, you need a nice programming guide. Look no further: Clicker has the solution.

Cable companies, be afraid. Internet TV is looming.

How broad is your band in the land?

2009-10-01T14:34:00.001+01:00

akamai's view

http://www.akamai.com/dv5

doesn't appear to be consistent with cisco's:

http://news.bbc.co.uk/1/hi/technology/8282839.stm

(e-mail by Jon Crowcroft)

SIGCOMM Day 4

2009-09-10T11:08:00.002+01:00

Session 6: Routing and Forwarding (Chair: Jia Wang, AT&T Research)

Stable and Flexible iBGP

Ashley Flavel (University of Adelaide), Matthew Roughan (University of Adelaide)

· Routing oscillation can decrease performance and lead to a high level of update churn.

· iBGP has also been shown to oscillate.

· More than eight years of research has not solved the problem of iBGP oscillation.

· Various solutions have been proposed but they all have problems: complicated to implement, restrict routing flexibility, or lack guarantees of stability.

· Their proposal is a very simple adaptation to the BGP decision process.

· They prove algebraically that it prevents iBGP oscillation.

LIPSIN: Line speed Publish/Subscribe Inter-Networking

Petri Jokela (Ericsson), Andras Zahemszky (Ericsson), Somaya Arianfar (Ericsson), Pekka Nikander (Ericsson), Christian Esteve (University of Campinas)

· Most Internet applications are internally publish/subscribe in nature

· Supporting efficient publish/subscribe requires 1) data-oriented naming, 2) efficient multicast, and 3) in-network caching.

· Deployment of native IP-based multicast has failed, and overlay- based multicast systems are inherently inefficient.

· Scalable and efficient publish/subscribe will require substantial architectural changes, such as moving from endpoint-oriented systems to information-centric architectures.

· Proposal: a novel multicast forwarding fabric, suitable for large-scale topic-based publish/subscribe. Fabric more energy efficient than the currently used ones.

PLUG: Flexible lookup modules for rapiddeployment of new protocols in high-speed routers

Lorenzo De Carli (University of Wisconsin-Madison), Yi Pan (University of Wisconsin-Madison), Amit Kumar (University of Wisconsin-Madison), Cristian Estan (University of Wisconsin-Madison), Karthikeyan Sankaralingam (University of Wisconsin-Madison)

· New protocols for the data link and network layer are being proposed to address limitations of current protocols (scalability, security, manageability).

· High-speed routers and switches that implement these protocols traditionally perform packet processing using ASICs. Why? High speed, low power, low chip area. But inflexible.

· Proposal: a flexible lookup module, PLUG (Pipelined Lookup Grid).

· They show that PLUG presents similar results to that of specialized lookup modules (in terms of throughput, area, power, and latency).

Session 7: Network Management (Chair: Thomas Karagiannis, Microsoft Research)

Modeling and Understanding End-to-End Class of Service Policies in Operational Networks

Yu-Wei Eric Sung (Purdue University), Carsten Lund (AT&T Labs Research), Mark Lyn (AT&T Labs), Sanjay Rao (Purdue University), Subhabrata Sen (AT&T Labs Research)

· Extensive use of service differentiation in Virtual Private Networks (VPNs) operated for business enterprises.

· The resulting Class of Service (CoS) designs embed complex policy decisions (bandwidth availability, cost).

· These complex high-level policies are realized through low-level router configurations (tedious process, error-prone ).

· They have proposed a formal approach to modeling CoS policies from router configuration files.

· They also present a practical and computationally efficient tool that can determine the CoS treatments received by an arbitrary set of flows across multiple routers

· They have validated their system using router configuration data from a cross-section of 150 diverse enterprise VPNs.

Towards Automated Performance Diagnosis in a Large IPTV Network

Ajay Mahimkar (The University of Texas at Austin), Zihui Ge (AT&T Labs - Research), Aman Shaikh (AT&T Labs - Research), Jia Wang (AT&T Labs - Research), Jennifer Yates (AT&T Labs - Research), Yin Zhang (The University of Texas at Austin), Qi Zhao (AT&T Labs - Research)

· IPTV is increasingly being deployed and offered as a commercial service.

· IPTV distribution network: hierarchical structure (instead of mesh), stringent requirements on reliability and performance, different distribution protocols (IP multicast), serious scalability challenges in managing millions of network elements.

· Characterisation and troubleshoot performance issues in one of the largest IPTV networks in North America.

· Measurement data used: device usage, error logs, user activity logs, video quality alarms, customer trouble tickets.

· They developed a novel diagnosis tool called Giza tailored to the enormous scale and hierarchical structure of the IPTV network.

· Multi-resolution data analysis to detect and localize regions experiencing problems.

· They make use of several statistical data mining techniques to troubleshoot the identified problems and diagnose their root causes.

Detailed Diagnosis in Enterprise Networks

Srikanth Kandula (Microsoft Research), Ratul Mahajan (Microsoft Research), Patrick Verkaik (UC San Diego), Sharad agarwal (Microsoft Research), Jitu Padhye (Microsoft Research), Paramvir Bahl (Microsoft Research)

· They analysed trouble tickets from small enterprise networks, and concluded that their operators need detailed fault diagnosis.

· They built a system (NetMedic) that enables detailed diagnosis by looking at info from operating systems and applications.

· Main idea: use the joint behaviour of two components in the past to estimate the likelihood of them impacting one another in the present.

· Prototype is effective at diagnosing faults that we inject in a live environment.

Session 8: Network Measurement (Chair: Gianluca Iannaccone, Intel Labs Berkeley)

Every Microsecond Counts: Tracking Fine-Grain Latencies with a Lossy Difference Aggregator

Ramana Rao Kompella (Purdue University), Kirill Levchenko (UC San Diego), Alex C. Snoeren (UC San Diego), George Varghese (UC San Diego)

· Network applications have stringent end-to-end latency requirements

· Fine-grain measurement demands cannot be met effectively by existing technologies, such as SNMP, NetFlow, or active probing.

· They propose a hash-based primitive LDA (Lossy Difference Aggregator) to measure latencies down to tens of microseconds and losses as infrequent as one in a million.

· They define a compact data structure that computes the average and standard deviation of latency and loss rate in a coordinated streaming environment.

· The system was compared with Poisson-spaced active probing with similar overheads. The LDA mechanism delivers orders of magnitude smaller relative error.

· Active probing requires 60 times as much bandwidth to deliver similar levels of accuracy.

Spatio-Temporal Compressive Sensing and Internet Traffic Matrices

Yin Zhang (University of Texas at Austin), Matthew Roughan (University of Adelaide), Walter Willinger (AT&T Labs -- Research), Lili Qiu (University of Texas at Austin)

· How to fill missing valies in a matrix (coulbe be a traffic matrix, delay matrix, social proximity matrix)

· They focus on traffic matrices.

· Missing values are quite common – direct measurement in infeasible, measurement unreliable, anomalies appear, and future traffic has not yet appeard

· Many network tasks are sensitive to missing values

· Ideas: 1) Exploit low rank nature of Traffic Matrices (TM); 2) exploit spatio-temporal properties (TM rows and columns close to each other are often close in value); 3) Exploit local strucytures in TMs

· Compressive sensing is used (generic technique for dealing with missing values that exploits the presence of structure and redundancy). However, existing compressive-sensing solutions perform poorly for traffic matrix interpolation.

· In the paper a novel spatio-temporal compressive sensing framework was developed. Key components: 1) a new technique called Sparsity Regularized Matrix Factorization (SRMF) that leverages the low-rank nature of traffic matrices and their spatio-temporal properties; 2) a mechanism for combining low-rank approximations with local interpolation procedures.

· They used real datasets from 3 networks.

· Compared to all other algorithms, this one is always better. Even with 98% missing values, the error is only of about 20%.

Passive Aggressive Measurement with MGRP

Pavlos Papageorgiou (University of Maryland, College Park), Justin McCann (University of Maryland, College Park), Michael Hicks (University of Maryland, College Park)

· Active proboing is very expensive – bandwidth expensive, and probes can interfere with the data

· With custom active measurement you shape the application data for measurement. This is efficient but not modular. And it is not reusable

· They propose MGRP – main intuition: MGRP piggybacks application data inside active probes.

· MGRP properties: 1) end to end measurement architecture (schedules probes for tx, and piggybacks application data on probes), 2) transparent to applications, 3) independent of measurement algorithms, 4) easy to adapt existing measurement tools, 5) can piggyback data across applications

· It enables aggressive probing with passive-like overhead.

· Piggybacking reduces bandwidth wasted by probes and enables measurement tools to be more aggressive faster and more accurate.

· Any measurement algorithm can now be written as if active, but implemented as passive.

Session 9: Performance Optimization (Chair: Ratul Mahajan, Microsoft Research)

ROAR: Increasing the Flexibility and Performance of Distributed Search

Costin Raiciu (UCL), Felipe Huici (UCL, NEC Labs), Mark Handley (UCL), David S. Rosenblum (UCL)

· We rely in distributed search everyday.

· Characteristics : big data set that doesn’t fit in one server. To reduce latency you use multiple servers, and put a bit of the data in each. But we also replicate the data. We create clusters of servers.

· PR = N (Number of Query Partitions * Replication = Number of Servers)

· P affects system behaviour. It dictates how much data each node stores, and it impacts latency.

· Partitioning determines latency and cost.

· Question: how to choose P, without knowing the workload? Choose the biggest value? Not desirable. CPU load increases...

· P is very difficult to change with existing solutions.

· How does google change P? In a way that requires over-provisioning, and lots of copies of the data (an estimate of 20TB/date center).

· The proposal: ROAR. Key observation: we do not need clusters to ensure each query meets all the data.

· ROAR uses consistent hashing. It uses a single parameter: P.

· ROAR copies zero data to increase P (while google’s algorithm has to copy a lot of data)

· Minimal data is copied when P decreases.

· They demonstrate the system using a privacy-preserving search application built upon ROAR.

· ROAR Changes P efficiently.

· The system scales. They tested this with 1000 servers in Amazon EC2.

Safe and Effective Fine-grained TCP Retransmissions for Datacenter Communication

Vijay Vasudevan (Carnegie Mellon University), Amar Phanishayee (Carnegie Mellon University), Hiral Shah (Carnegie Mellon University), Elie Krevat (Carnegie Mellon University), David Andersen (Carnegie Mellon University), Greg Ganger (Carnegie Mellon University), Garth Gibson (Carnegie Mellon University and Panasas, Inc), Brian Mueller (Panasas, Inc.)

· Timeout delay in TCP is big – 200 ms could be a big problem for datacenters.

· Some datacenter applications are very sensitive to these 200ms timeouts.

· Solution: enable microsecond retx (being safe in the wide area).

· Datacenter environment: 1-10Gbps, commodity Ethernet switches, and 10-100 microseconds latency.

· In a datacenter environment, 1 TCP timeout means more than 1000 RTTs...

· Long periods of idle links due to these high timeouts

· Other TCP variants did not prevent TCP timeouts.

· Timeouts also increase latency

· The solution: Microsecond TCP retransmissions

· 1^st solution: simple one line change in Linux (eliminate minRTO). It improves throughput, but still not microsecond RTO.

· Requirements for microsecond RTO: TCP must track RTT in microseconds (timestamp option). Efficient high resolution kernel timers (use HPET for efficient interrupt signalling).

· They were able to provide high throughput sustained for up to 47 servers.

· Is it safe? One has to look at the interaction with delayed ACK, and the performance in the wide area.

· Better performance with delayed ACK disabled.

· Is it safe for the wide area? Potential concerns: stability (could we cause congestion collapse?), and performance (do we oftenb timeeout unnecessarily?)

· Stability is preserved, because timeouts retain exponential backoff and spurious timeouts slow rate of transfer.

· Today’s TCP has mechanisms to detect spurious timeouts (using timestamp) and to recover from spurious timeour (forward RTO), and they are both implemented widely.

· They analysed wide area performance without minRTO, to see if it harms the wide area. They didn’t find any difference in throughput. Reason? Few total timeouts (supuious or legitimate).

Matchmaking for Online Games and Other Latency-Sensitive P2P Systems

Sharad Agarwal (Microsoft Research), Jacob R. Lorch (Microsoft Research)

· Latency is the main factor to be aware of in multiplayer online games.

· Predicting latency would help.

· They have an extensive matchmaking trace.

· Current approaches: geolocation, and network coordinate system. They mix both to get Htrae.

· Weakness of geolocation: inflexible. Weakness of network coordinate systems: sensitive to initial conditions (takes a while to converge).

· They have refined their original system that basically mixed geolocation for the initial condition and a network coordinate system to have flexibility. Refinements: 1) autonomous system correction (to change the height of the node); 2) symmetric updates to speed up convergence.

· Trace replays for evaluation: 30 days, 3.5 million machines, 50 million probes.

· Absolute error median for Htrae is 15ms.

· Htrae can find the best server 70% of the time.

· Deployed systems have to guess a lot, so Htrae is way better.

SIGCOMM Day 3

2009-08-24T23:26:00.004+01:00

Session 3: Network Architecture (Chair: Renata Teixeira, LIP6)

De-anonymizing the Internet Using Unreliable IDs

Yinglian Xie (Microsoft Research Silicon Valley), Fang Yu (Microsoft Research Silicon Valley), Martín Abadi (Microsoft Research Silicon Valley and UCSC)

· The Internet is open and anonymous. Therefore, attackers that generate malicious traffic cannot typically be held accountable.

· HostTracker is presented. It tracks dynamic bindings between hosts and IP addresses by leveraging application-level data with unreliable IDs.

· They use a month-long user login trace from a large email provider

· HostTracker can attribute most of the activities reliably to the responsible hosts, despite the existence of dynamic IP addresses, proxies, and NATs.

SmartRE: An Architecture for Coordinated Network-wide Redundancy Elimination

Ashok Anand (University of Wisconsin-Madison), Vyas Sekar (Carnegie Mellon University), Aditya Akella (University of Wisconsin-Madison)

· Application-independent Redundancy Elimination (RE) (identifying and removing repeated content from network transfers), is used to improve network performance.

· A network-wide RE service would be beneficial for ISPs (to reduce link loads, increase effective network capacity)

· The authors present SmartRE, a architecture for network-wide RE.

· SmartRE enables more effective utilization of the available resources at network devices.

· They used real and synthetic traces to evaluate.

Design and Implementation of High Performance Dual-radio Mesh Networks

Aditya Dhananjay (New York University), Jinyang Li (New York University), Lakshminarayanan Subramanian (New York University), Hui Zhang (Tsinghua University)

· How to realise the full potential of a multi-radio mesh network? 1) how to assign channels to radios at each node to minimize interference; 2) how to choose high throughput routing paths in the face of lossy links, variable channel conditions and external load?

· ROMA is a distributed channel assignment and routing protocol that achieves good multi-hop path performance.

· They assign non-overlapping channels to links along each gateway path to eliminate intra-path interference.

· They reduce inter-path interference by assigning different channels to paths destined for different gateways whenever possible.

· They evaluated on a 24-node dual-radio testbed.

Session 4: Novel Aspects to Networking (Chair: Jon Crowcroft, Cambridge University)

Pathlet Routing

P. Brighten Godfrey (University of Illinois at Urbana-Champaign), Igor Ganichev (UC Berkeley), Scott Shenker (ICSI and UC Berkeley), Ion Stoica (UC Berkeley)

· Pathlet is a new routing protocol.

· Networks advertise fragments of paths (pathlets) that sources concatenate into end-to-end source routes.

· Pathlet routing can emulate the policies of BGP, source routing, and several recent multipath proposals.

· When a router's routing policy has only local constraints, it can be represented using a small number of pathlets, leading to very small forwarding tables

· Pathlet does not impose a global requirement on what style of policy is used, but rather allows multiple styles to coexist.

Cutting the Electric Bill for Internet-Scale Systems

Asfandyar Qureshi (MIT), Hari Balakrishnan (MIT), John Guttag (MIT), Bruce Maggs (Akamai/CMU), Rick Weber (Akamai)

· Energy expenses are becoming an increasingly important fraction of data center operating costs.

· Energy expense per unit of computation can vary significantly between two different locations.

· The paper characterizes the variation due to fluctuating electricity prices.

· Existing distributed systems should be able to exploit this variation for significant economic gains.

· Electricity prices exhibit both temporal and geographic variation.

· They use simulation to quantify the possible economic gains for a realistic workload.

· Existing systems may be able to save millions of dollars a year in electricity costs.

Persona: An Online Social Network with User-Defined Privacy

Randolph Baden University of Maryland Adam Bender (University of Maryland), Daniel Starin (University of Maryland), Neil Spring (University of Maryland), Bobby Bhattacharjee (University of Maryland)

· In OSNs users share private content, and trust the OSN service.

· Persona is an OSN where users dictate who may access their information.

· They hide user data with attribute-based encryption (ABE), allowing users to apply fine-grained policies over who may view their data

· They describe an implementation of Persona that replicates Facebook applications and show that Persona provides acceptable performance.

Session 5: Wireless Networking 2 (Chair: Suman Banerjee, University of Wisconsin at Madison)

In Defense of Wireless Carrier Sense

Micah Z. Brodsky (MIT), Robert T. Morris (MIT)

· The wireless medium is semi shared

· Carrier Sense: “Can I talk now?”. Interference protection and space reuse. Very simple.

· Is it too simple? If networks are far apart, concurrency is the best option. If they are close, time mux. What about in the middle?

· Main question: How well does CS work?

· When does CS works well? When interferer is very far away, or when it is very close to the sender. Intermediate distance is the hard case. What about shadows and obstacles?

· They start with a simple model, only 2 contending tx, with same power, omni antennas

· The effect of varying sender-sender distance is investigated.

· They use a standard model for radio propagation that include path loss and environmental shadowing. They ignore multipath fading because wideband channels average this away.

· They use Shannon capacity as a model for throughput (Adaptive bit rate)

· Answer for intermediate problem: Adaptive Bit Rate.

· Obstacles aren’t fatal. Usually you have alternate propagation paths.

· By analysing average throughput, they realise Carrier Sense works.

· Intuitions summary: Distant interferers affect receivers uniformly; nearby interferes don’t but they’re loud so everybody prefers mux anyway; rate adaptation helps in intermediate situation; and shadowing is not such a big problem.

· Implications for future research: adaptive bit rate is essential.; hidden terminals can be a problem in terms of reliability, but they don’t matter much for average performance; exposed terminals don’t cost very much, if ABR is working.

· Carrier sense does work. There is room for improvement, but not much in overall performance.

Interference Alignment and Cancellation

Shyamnath Gollakota (MIT), Samuel D. Perli (MIT), Dina Katabi (MIT)

· MIMO LANs increase throughput by sending more concurrent packets.

· In this paper the authors present a technique that doubles concurrent packets in MIMO LAN

· Concurrent MIMO decodes as many concurrent packets as there are antennas per AP. Can we do better?

· With 2 antennas, current MIMO LANs can decode only 2 packets. All current MIMO LANs are limited by number of antennas per AP.

· What if the APS coordinate over the Ethernet? 2 APs with 2 antennas each could communicate via Ethernet, and then decode more than 2 packets. But there is an impractical overhead in this solution. Can we leverage the Ethernet with minimal overhead?

· Their solution: Interference Alignment and Cancelation (IAC)

· IAC overcomes the antennas per AP throughput limit. A packet is decoded then broadcasted once on the Ethernet, with minimal overhead.

· Contributions of the work: first MIMO LAN to overcome the antennas per AP limit; IAM synthesises interference alignment and cancelation; IAM doubles MIMO throughput; implementation of the scheme in software radios to prove this.

· For a large number of antennas, IAC doubles MIMO throughput

· They tested with a 20 node testbed. Uplink gain: IACs median gain is 2.1x better than current MIMOs. Gain is partially due to diversity but even more to concurrency. Downlink: IAC median gain is 1.5x. IAC is beneficial across the operational range of SNRs.

DIRC: Increasing Indoor Wireless Capacity Using Directional Antennas

Xi Liu (Carnegie Mellon University), Anmol Sheth (Intel Research Seattle), Michael Kaminsky (Intel Research Pittsburgh), Konstantina Papagiannaki (Intel Research Pittsburgh), Srinivasan Seshan (Carnegie Mellon University), Peter Steenkiste (Carnegie Mellon University)

· Driving demand for wireless capacity. Interference can be a big issue.

· Goal: use directional antennas to improve wireless capacity by increasing spatial reuse.

· They use phased array antennas – they electronically steer the signal to a specific direction, hence having small reconfiguration time. They assume that only APs use these antennas – they are too bulky for clients.~

· Limitations in indoor environment: LOS may be blocked, indoor space is rich scattered. The conventional wisdom is that directional antennas are not effective in indoors environment.

· Key idea: leverage multiple paths and obstacles to improve spatial reuse

· How to find antenna orientations? 1) Naive solution, max cap, is too slow. The objective is to find optimal antenna orientations quickly.

· How to coordinate between antennas? They use a centralised controller, and TDMA scheduling MAC. They also separate directional and omni-directional antennas.

· They made measurements in 2 indoor environments: 3 directional APs and 6 omni clients in each testbed.

· 2x improvement over OMNI CSMA and 1.6x over MAX SNR.

· Conclusion: coordination is required to use directional antennas effectively in indoors environments

SIGCOMM Day 2

2009-08-19T14:51:00.002+01:00

Keynote Speech

Great keynote speech by the winner of the SIGCOMM award (well deserved!): Jon Crowcroft. Here are the slides:

http://conferences.sigcomm.org/sigcomm/2009/pecha-kucha-dozen.pdf

Session 1: Wireless Networking 1 (Chair: Brad Karp, University College London)

Cross-Layer Wireless Bit Rate Adaptation

Mythili Vutukuru (MIT), Hari Balakrishnan (MIT), Kyle Jamieson (UCL)

· We have time varying wireless channels: due to large scale attenuation, small scale fading and interference

· So we need online bit rate adaptation: varying modulation and coding.

· Currently we have frame-based and SNR based algorithms for this.

· They have problems: slow, need look up tables, so they propose SoftRate. Use per-bit confidences, no need for look up tables – they get interference free BER.

· SoftPHY design more general than earlier work.

· Adapts to channel accurately and quickly, robust to collision losses, 2x gains over existing protocols.

· They propose using a soft output decoder (instead of the normal decoder) in the receiver, and use a different protocol, SoftRate.

· They created a GNU radio with USERP. Physical layer was from real traces, but then used ns-3 for TCP. They used a channel simulator for some scenarios (like train travelling).

· Good results predicting the BER of the channel.

· The comparison was made with other protocols: static best (best for each packet), SNR-based and Frame based.

· Compared to the optimum (static best): was within 10% of the optimal.

· Compared to the frame based: up to 2x over best frame based (these are very slow).

· Compared to SNR based: 4x over untrained SNT based algorithms

SMACK - A SMart ACKnowledgment Scheme for Broadcast Messages in Wireless Networks

Aveek Dutta (University of Colorado at Boulder), Dola Saha (University of Colorado at Boulder), Dirk Grunwald (University of Colorado at Boulder), Douglas Sicker (University of Colorado at Boulder)

· Question: Can we reduce the ACK time for broadcast/multicast scenarios?

· Instead of each user answering at its time, multiple users response at the same time to reduce the ACK time – using OFDM.

· The objective is to speed up group communication, like route discovery, neighbour info, etc.

· Nodes are assigned unique sub carriers. They send a tone to say “yes”.

· No packet transmission + concurrent response = faster ACK

· They have made an implementation of the system.

· In summary, main idea: PHY layer signalling can be used to innovate new protocols for wireless networks.

White Space Networking with Wi-Fi like Connectivity – Best paper award

Paramvir Bahl (Microsoft Research), Ranveer Chandra (Microsoft Research), Thomas Moscibroda (Microsoft Research), Rohan Murty (Harvard University), Matt Welsh (Harvard University)

· Main objective: How to build a wireless network using the white spaces?

· Spectrum allocation: there is more spectrum for broadcast TV than to WiFi.

· Moving from analog TV to digital TV.

· White spaces: unoccupied TV channels. Let's use them!

· We must not interfere with TV and mikes that are using that part of the spectrum: so we can use it iff no one else is using it.

· So we have more spectrum (3x that of 802.11g) and longer range (at least 3 to 4x WiFi)

· Goal: deploy infrastructure wireless – give good throughput without interfering with incumbents (TV and mike)

· Problems: fragmentation of spectrum (so we have variable channel widths), location impacts spectrum availability (spectrum exhibits spatial variation), and there is also temporal variation (incumbents appear/disappear over time).

· They’ve built the WhiteFi system – to evaluate by deployment of prototypes and by simulations.

· How do the new clients know which channels to use (discovery time)? They infer by analysing for how long the amplitude of a received signal is increased. They achieve a 2x reduction of discovery time for 30MHz width.

· Spectrum assignment algorithm: they implement MCHAM – a multi channel airtime metric. They consider not only if the channel has room, but also how much it has.

Session 2: Datacenter Network Design (Chair: Stefan Saroiu, Microsoft Research)

PortLand: A Scalable Fault-Tolerant Layer 2 Data Center Network Fabric

Radhika Niranjan Mysore (University of California San Diego), Andreas Pamboris (University of California San Diego), Nathan Farrington (University of California San Diego), Nelson Huang (University of California San Diego), Pardis Miri (University of California San Diego), Sivasankar Radhakrishnan (University of California San Diego), Vikram Subramanya (University of California San Diego), Amin Vahdat (University of California San Diego)

· Portland is a single logical layer 2 data centre network fabric. It separates host identity (IP) with host location (a “PMAC”).

· Data centres are growing in scale

· Goals for data centre network fabrics: plug and play, scalability, small switch state, seamless VM migration

· Layer 2 data centre fabrics. Advantages: plug and play, and seamless VM migration.

· Layer 3 data centre fabrics. Advantages: scalability, small switch state.

· With flat address you need about 100MB of chip memory (150 times what we put in a chip today).

· Other network fabrics: SEATTLE (SIGCOMM08) – problems: large switch table and broadcast based routing protocol. VL2 (SIGCOMM09).

· Portland: Plug and Play + Small Switch state.

· Main assumption: Hierarchical structure of data centre networks; they are multilevel, multi-routed tree.

· They impose a hierarchy on a multi-rooted tree.

VL2: A Scalable and Flexible Data Center Network

Albert Greenberg (Microsoft Research), Navendu Jain (Microsoft Research), Srikanth Kandula (Microsoft Research), Changhoon Kim (Princeton), Parantap Lahiri (Microsoft Research), David A. Maltz (Microsoft Research), Parveen Patel (Microsoft Research), Sudipta Sengupta (Microsoft Research)

· Cloud service data centre need to be agile (assign any servers to any services) and must scale out.

· VL2: First DC network that enables agility in a scaled out fashion.

· They analysed a large cluster, and realised that traffic patterns are highly volatile and unpredictable – so optimisation should be made frequently and rapidly

· We need a huge L2 switch, or an abstraction of one

· VL2 achieves agility at scale via 1) L2 semantics, 2) uniform high capacity between server, and 3) performance isolation between services

· Lessons: 1) randomisation can tame volatility, 2) add functionality where you have control, 3) there’s no need to wait.

BCube: A High Performance, Server-centric Network Architecture for Modular Data Centers

Chuanxiong Guo (Microsoft Research Asia), Guohan Lu (Microsoft Research Asia), Dan Li (Microsoft Research Asia), Haitao Wu (Microsoft Research Asia), Xuan Zhang (Tsinghua University), Yunfeng Shi (Peking University), Chen Tian (Huazhong Universtiy of Science and Technology), Yongguang Zhang (Microsoft Research Asia), Songwu Lu (UCLA)

· Novel network architecture for container based, modular data centres.

· BCube design goals: high network capacity for various traffic patterns (one to one unicast, one to all and one to several groupcast, and all to all data shuffling); only use low-end commodity switches, graceful performance degradation

· BCube is a server centric network.

· They compare their system with Tree, Fat-Tree, and DCell+, achieving higher performances.

SIGCOMM, Day 1

2009-08-18T16:15:00.002+01:00

MobiHeld Session III: Services

Chair: Lakshminarayanan Subramanian (New York University)

Virtual Individual Servers as Privacy-Preserving Proxies for Mobile Devices

Ramón Cáceres (AT&T Labs), Landon Cox (Duke University), Harold Lim (Duke University), Amre Shakimov (Duke University), Alexander Varshavsky (AT&T Labs)

· Main goal: keeping ownership and control of your data

· Idea: each person has its own virtual machine

· People increasingly upload content from their mobile devices to 3^rd party services (facebook, twitter, etc.)

· This leads to privacy issues. They focus on 2 issues: 1) these services are centralised (vulnerable to large scale privacy breaches), 2) terms of service often grant provider rights to user data

· Virtual Individual Servers: instead of uploading our info to 3^rd party services, upload data to a VIS (a machine the user owns). Individuals maintain rights to their data. Data is distributed across many administrative domains.

· Advantages: privacy, flexibility (my own machine, I can install whatever I like), long term availability, cost scalability

· Disadvantages: management burden (users are bad to manage their machines at home, so managing a virtual machine will be complicated), cost to the individual

· VISs vs. serving data from devices – advantages: resource richness, high availability; disadvantages: requires access to wired infrastructure, need network connection

D^3N: Programming Distributed Computation in Pocket Switched Networks

Eiko Yoneki (University of Cambridge), Ioannis Baltopoulos (University of Cambridge), Jon Crowcroft (University of Cambridge)

· Evolution of mobile networks: a more disconnected network: a path from A to B may exist, but only over time.

· Looking at human to human connectivity

· Use of declarative networking

Apprehending Joule Thieves with Cinder

Stephen M. Rumble (Stanford University), Ryan Stutsman (Stanford University), Phil Levis (Stanford University), David Mazieres (Stanford University), Nickolai Zeldovich (MIT)

· Desktop resource management: if it’s slow, add more resources

· State of mobile devices: complex... and users care about energy and network

· Future of mobile devices: need new OS mechanisms for resource management

· Consider energy as a first class resource: track it, ration it, delegate it.

· They define a “capacitor abstraction” to explain the way they manage the mobile phone energy usage –kind of a leaky bucket concept.

· Capacitors can offer fine grained tracking, rationing and delegation. They easily express real world policies.

Game Action Based Power Management for Multiplayer Online Game

Bhojan Anand (National University of Singapore), A.L. Ananda (National University of Singapore), Mun Choon Chan (National University of Singapore), Rajesh Krishna Balan (Singapore Management University), Le Thanh Long (National University of Singapore)

· Main contribution: game action based resource management

· Black box approach: Lose some packets to save energy; White box approach: reduce number of packets, use some AI to remove redundancy – both baseline approaches failed.

· Application assisted approaches: go to off or deep sleep mode, without reducing quality

· Check Player Activity Level (PAL) – if it’s low, go to sleep mode.

· Can we predict current PAL with the previous PALs? Yep.

· How long can we put the WNIC to sleep? Must find optimum.

· They also looked at the frequency of game actions (shooting, walking, etc.)

Optical Communications and Networking

2009-06-18T22:50:00.001+01:00

For those interested in Optical Networking, the inaugural issue of the Journal of Optical Communications and Networking (JOCN) is now available:

http://www.opticsinfobase.org/jocn/Issue.cfm?utm_source=memnews&utm_medium=Email&utm_campaign=0906

JOCN is the result of a partnership that merges JON (Journal of Optical Networking) with ComSoc's Supplement on Optical Networking and Communications.

Tenure and the Future of the University

2009-06-04T21:36:00.001+01:00

An interesting article on the tenure system:

http://www.sciencemag.org/cgi/content/full/324/5931/1147?rss=1

Top journals in computer science

2009-05-28T22:45:00.000+01:00

The most recent list here in the Times Higher Education.

OFC 2009, Day 5 (The End)

2009-03-27T02:30:00.002Z

“Network Coding and Its Implications on Optical Networking”

This was an invited talk by Muriel Médard, from the MIT. Some notes:

1. Multicast capacity can be achieved with network coding. Minimum cost multicast is possible by allowing coding.

2. What about achieving the network coding advantage by coding only a subset of nodes? Cost of doing in every node would be prohibitive…

3. Some interesting paper on this topic: “Information-theoretic framework for network management for recovery from non-ergodic link failures” (Ho, Medard, etc.), “Concept of 1*N protection that uses network coding over p cycles”, Kamal 2006

4. Other interesting paper: Menendez and Gannett, 2008, show that network coding can lead to significant savings in back up resources by allowing coding. Only bitwise XOR is considered as it can be handled in the optical domain (in the electronic domain we don’t have this limitation, but the problem is that we need OEO conversion…).

5. Her group has showed that it’s possible to have benefits of coding even without having all nodes doing it.

6. Finally another paper with an evolutionary approach (Infocom 07) to minimise coding cost (number of coded lambdas).

All this network coding stuff could be of interest for multicast…

“CAPEX and OPEX in Aggregation and Core Networks”

This was yet another invited talk by Claus Gruber, Nokia. Some notes:

1. From the slides: “Video services will drive exponential growth… especially from IPTV”.

2. Revenues and traffic were usually correlated, but they are now being more decoupled. So we have maintained the network costs low… otherwise were done.

3. How to reduce CAPEX? Some examples: a) transport bits on the lowest level – it’s cheaper, b) the network should be automated, by using GMLS as control plane and path computation elements, c) use virtualisation techniques to enable fast cost efficient deployment of new services, d) optimal mix of intermediate grooming and routing and transport bypass.

4. How to reduce OPEX? Reduce energy consumption. Internet technologies represent 1 to 4% of energy consumption today…

5. IP/MPLS routers consume more power (1kW /100Gbps). Next in the list L2 switches, followed by SDH nodes, and finally ROADM nodes…

6. Extend traffic engineering decisions with energy profile info. Try to find paper “Energy profile aware routing”. I think this guy stole my idea in this paper… OK, he was first! :) But making routing “green” seems a good thing, and it’s a hot topic definitely.

“Architectures for Energy-Efficient IPTV Networks”

Jayant Baliga presented a paper with a new energy consumption model of IPTV storage and distribution (IPTV in his talk was VoD). Some notes:

1. Energy becoming an issue – OPEX, greenhouse gases. He focused on video delivery over public internet.

2. Some notes on power consumption: ethernet switch – 9 nJ/b, Broadband net gw (140 nJ/b), router (26 nJ/b). Core net: core router (17 nj/b), optical OXC (0.02 nJ/b). Data center: edge router (26 nJ/b), server (430nJ/b). Etc.

3. Packets traverse an average of 13 hops (3 hops metro + 9 on core + 1 data centre).

4. They check how many data centres they should use. The conclusion is that unpopular movies would stay in few data centers, say 1 or 2, but a very popular one would stay in several – he said 200 was a good number.

5. Optical bypass reduces power consumption.

6. P2P is efficient at low downloads per hour (unpopular content), but for high demand content is not efficient at all (due to upstream speeds being low).

“Power Saving Architectures for Unidirectional WDM Rings”

Another paper by our colleagues from Italy, presented by Piero Castoldi. The paper compares the optimal power-saving designs of unidirectional WDM rings using three well-known architectures, first-generation, all-optical, and multi-hop (hybrid). Notes:

1. Are multihop architectures able to offer a reduction of OPEX (they reduce CAPEX)?

2. They have analysed power savings… and used ILP for this. These are the values they used (from work published recently): power of electronic interfaces = 197mW, power of optical interfaces = 189mW.

Main problem: it is an ILP formulation, so not scalable. Heuristics needed… maybe we can collaborate with them again? :)

OFC 2009, Day 4

2009-03-27T01:52:00.001Z

“An Offline Impairment Aware RWA Algorithm with Dedicated Path Protection Consideration”

The first author of this paper was Siamak Azodolmolky. He presented a novel offline physical layer impairment aware routing and wavelength assignment algorithm for various transparent all-optical networks, while considering dedicated path protection. He started by presenting a network planning and operation tool they’ve developed, focusing his presentation on a specific part of it, the impairment aware lightpath routing component of the tool. They consider several impairments, like node crosstalk, PMD, XPM, FWM, etc. and calculate a Q-factor. This is only for offline settings (used for network planning), but in an online setting they would have to change the way the Q factor is currently calculated, because it is very computationally intensive today.

“Impact of Topology and Traffic on Physical Layer Monitoring in Transparent Networks”

This was an invited talk by Dan Kilper, Bell Labs, Alcatel-Lucent, USA. He analysed the benefit of optical performance monitoring, with respect to network topology and traffic patterns. So in OEO networks we can monitor things in the regenerator sites. But in optical nets this is more complicated. So the question is: where to place monitoring sites? They have tested some schemes, and shown that we can reduce costs significantly by placing monitoring sites in specific locations.

“Optical Multi Domain Routing”

This was also an invited talk. The speaker was Xavi Masip.

He reviewed the current limitations in multi-domain routing as well as some of the research lines in the optical area. Some notes:

1. Mixing routing, multi domain, and optics makes things very complex. Today it is known that all ISPs are being glued together with BGP. So to change it is very complicated – no one wants to lose connectivity. And BGP works.

2. Why multilayer? To optimise layers: we have one top IP layer, then a lower one for connection oriented packet networks, and finally the photonic transport layer. With multilayer optimisation we one can reduce costs and power needs.

3. Today we don’t have multidomain optical routing, BGP takes care of everything. There is some work on this area, but more research is needed.

4. Check work on Optical BGP (OBGP), and also OBGP+. Could be interesting.

5. Do we need multi domain optical routing? Answer by pure transport people (optical people): NO (upper layers can do that). Answer by IP layer people: definitely not (we can do that, just provide connectivity). Answer by carriers: are we supposed to share info?... (confidentiality is very important)

“PCE Communication Protocol for Resource Advertisement in Multi-Domain BGP-Based Networks”

The presenter was Francesco Peolucci. They propose new PCEP messages to announce inter-domain resource information typically not advertised by BGP because of scalability reasons. These enable effective PCE-based path computations and preserve network stability, scalability, and intra-domain information confidentiality. Some notes:

1. BGP doesn’t advertise alternative solutions. BGP applies tie breaking rules, and only advertises one route, no alternative paths – for scalability reasons.

2. Proposed solution – extend PCE protocol with one new message that carries route info that is not advertised by BGP. They also propose to include another message with bandwidth availability of interdomain traffic engineering.

3. This is limited to a set of domains (they consider only 20 domains), not to the entire internet… so no worries about scalability. But the fact is that I was still worried about scalability. They don’t address this, they only say “it is not a problem” because they limit themselves to 20 domains (not sure why).

“On Resource Provisioning for Dynamic Multi-Domain Networks”

Xiaolan J. Zhang gave an interesting talk where she studied the performance of multidomain resource dimensioning/routing techniques with limited information sharing, and provided motivation for considering fairness issues. Some notes:

1. Imagine your network is well dimensioned, but your “neighbour” network is underprovisioned. Your connections to that network are rejected because of them, so you don’t have a big incentive to have a well dimensioned net… One domain can hurt performance of other domains.

2. A global shortest path routing solution prefers larger domains – large domains have advantage over smaller domains.

3. So they propose a global SPF but they normalise the costs to the size of the network. They improve performance on smaller nets, without causing problems in the big network.

“Avoiding Path-Vectors in Multi-Domain Optical Networks”

Marcelo Yannuzzi showed that a modified version of a path vector protocol can drastically reduce the blocking and converge significantly faster, while exchanging less number of routingmessages both during failure-free conditions and during a convergence. More notes:

1. OBGP+: A modified Path Vector (ENAW), including a cost that is dependent on number of lambdas available.

2. OBGP+ performs way better than OBGP in terms of blocking, number of messages exchanged, convergence time in case of failure, routing advertisements (churn).

3. Conclusions: even minor modifications are sufficient to outperform a plain path vector like OBPP. Avoid plain path vector for the interdomain RWA protocol.

OFC 2009, Day 3

2009-03-25T19:53:00.002Z

This was a great day in OFC. Several very, very interesting talks. Now I start understanding why they say OFC is the top conference in optical comms… I went to the plenary session in the morning, with three excellent talks, and then in the afternoon I decided to go to the Symposium on the Future Internet. Also a couple of nice talks there.

Plenary Session

"The Growth of Fiber Networks in India"

The speaker was Shri Kuldeep Goyal, Chairman and Managing Director of Bharat Sanchar Nigam Ltd. (BSNL), India (largest SP in India, Government owned). It was a very interesting presentation. The first slides were very interesting. He started by showing a world map. As in any "normal" world map, it is a geogrophical scale, i.e., bigger countries appear bigger in the map. Then he change the type of scale of the map, to his own convenience, which was really neat. For example, he showed the map with a population scale, and we could see India inflating hugely. Then he showed the same map with yet different scales: broadband connections (India was very thin in here), films watched (India was more inflated in here than in the population map), etc.

Some notes:

1. BSNL is offering IPTV in 30 cities today. It has 25k users at the moment.

2. They are now starting to move to FTTH.

3. They offer 610 TV channels!

4. In January 2009 there were 15 million new users of mobile phone in India.

5. They use IP/MPLS in their core network.

6. 85% of the broadband technology in India is ADSL.

7. He expects that, by 2011, 5% of all households will have FTTH.

8. For him IPTV is one of the main drivers of FTTH.

9. In Japan the already have 25% of the households with FTTH.

10. He wants to offer HDTV soon.

“The Changing Landscape in Optical Communications”

This was a very dynamic and also very interesting talk by Philippe Morin, the President of Metro Ethernet Networks, Nortel. Some notes:

1. One of the three “global megatrends” he mentioned in his talk was HDTV. He believes the demands for HDTV video are fuelling the next wave of bandwidth growth.

2. Talking about business applications, he things the top service will be telepresence/videoconference (to reduce costs by decreasing the number of flights).

3. He mentioned Oprah’s “A New Earth” classes. More than 1 million people was watching this online, simultaneously, from 133 different countries, HQ 1.5 Mbps. This really is asking for multicast… :)

4. Another example was Obama’s inauguration, with millions of people watching simultaneously. Real time is definitely needed, in his (and my) opinion.

5. He says that anything less than HD will be unacceptable in 2 years time. And don’t forget 3DTV, doubling the bandwidth needs…

6. Another nice example of needs for live video was the NCAA championship that allows HQ streaming for the PC.

7. He showed a nice figure with the evolution of TDM, until 1995/200, then the evolution of WDM, from 2000 to 2010, and he thinks the next thing are the coherent techniques.

8. Next frontier: terabit per lambda, and access technology cost going down one order of magnitude.

“Getting the Network the World Needs”

The speaker was Lawrence Lessig, a Professor in Stanford Law School. I have to be honest: this was the best talk I have ever seen. It was absolutely amazing. Impressive. Superb. He was a Cambridge student (did his masters there), so I guess that explains the high quality a bit… lol :)

It is very difficult to explain his style, because it is very different from the things we are used to. But you can check an example of his style here (you can check others in youtube, it is really cool!):

Someone that is able to put this full video in the middle of the talk, is certainly delivering a great presentation:

By the way, with all the excitement I almost forgot... I didn’t take notes because I was completely absorbed by the presentation, but I can just say that the main topic of the talk was this: Copyright needs change.

Again, superb.

Now moving to the Symposium on the Future Internet…

“Future Internet: Drastic Change, or Muddling Through?”

The speaker was Professor Andrew Odlyzko, University of Minnesota. This was basically the same talk Professor Odlyzko gave some months ago in Cambridge, so it was not new for me. Nevertheless, it is always interesting to hear him. And I leave some brief notes:

1. Huge potential sources of additional Internet traffic: a) storage, and b) broadcast TV. In 2006, in the US, internet traffic per capita was 2GB/month, but TV consumption per capita was 40 GB/month – assuming 3 hours/day, 1Mbps, no HDTV… so it’s a soft figure.

2. Net neutrality “is about streaming movies” (Jim Cicconi, AT&T, 2006)

“Internet Evolution into the Future”

Another great talk, this time by Lawrence Roberts, one of the fathers of the Internet (responsible for the first packet network, ARPANET), now Chairman of Anagram (he was also the founder of this company).

It is always exciting to listen one of the founding fathers of the Internet (some time ago I listened to a talk by Vincent Cerf and that was also exciting!). In 1965, in MIT, Roberts made the 1^st packet network experiment, and then he managed the first packet network, ARPANET. Some notes:

1. He showed a nice slide with the Original Internet Design, with its main activities: file transfer and email. At that time, only packet destination examined, no source checks, no QoS, no security, best effort, video not feasible. And guess what: Not much change since then!

2. But now, some changes: voice is moving to packets, video is moving to packets… Now edge is broad, and core is narrow!

3. P2P is a problem, because it uses TCP unfairness – multiple flows, so multiple capacity, and flows are treated equally (equal capacity per flow)… so it congests network, with 5% users using 80% capacity.

4. So he proposes changes to the Internet: 1) fairness (multiflows – p2p – applications overload network), 2) security (user auth, source check), 3) emergency services (secure preference priorities), 4) cost and power (make network green), 5) quality (video and voice require lower jitter and loss).

5. What we need? Equal capacity for equal pay.

5. Today all security is left to the computer. The network doesn’t even identify the source address (no way to determine who sent spam…). Goal: network to secure each connection (flow) – user and computer ID sent to network to be verified, and network check if source address is correct.

6. Other proposal: Flow rate management – interesting subject: control the rate of each flow individually, ensure congestion does not occur by controlling rate. Use rate control, instead of random drops.

7. Why flow rate management now? Memory cost has come down faster than processing cost – and flow rate management is memory based…

8. Cost and Power: today network equipment is packet based (every packet re-examined)… with flow rate management we process flows (just look at flow record). With flows (not packets), power, size and cost lowers 3 to 5 times…

“Building a zero carbon Internet”

A presentation by Bill St. Arnaud, from CANARIE, Canada. Some notes:

1. They are building “zero carbon” data centres connected by optical networks, powered by windmill and hydroelectric power, etc. These are built in remote locations.

2. Optical networks, especially 100G and 1000G, have modest increase in power consumption, way better than electronics.

3. Funding in UK will be linked to reduction of carbon emissions.

4. They are in a project, PROMPT, that tries to lower cost power by following “sun and wind”…

5. Research ideas: a) dynamic all optical nets with solar or wind powered optical repeaters, b) wireless mesh adhoc nets with mini solar panels at nodes, c) new shortest energy path internet architectures with servers, computers and storage collocated at remote renewable energy sites… This is a great topic, and everybody at OFC was talking about being “greener”. Maybe we could engineer traffic to make the network green? Now that would be a cool idea…

6. Check http://green-broadband.blogspot.com for more info.

OFC 2009, Day 2

2009-03-25T02:57:00.001Z

OFC is a very big conference, with several interesting things happening at the same time. Therefore, we have to be very judicious in the choices we make. Some of the talks are quite good, but many of them are really uninteresting, or the speakers are really poor, or are just very nervous. So I will only make notes of talks I think were worthy somehow... which will leave probably half out.

In this first day of technical presentations I seem to have perceived a trend: the optical community is trying to tie itself closer together with the internet community. There was a talk by someone from the National Science Foundation and also another from the GENI project where this was quite clear. But let me leave in here some details on a number of talks.

"Global Load Balancing of Zero-Bandwidth TE LSPs in MPLS Networks"

The first author of this paper is Filippo Cugini (from CNIT), but the presenter was Francesco Paolucci, from Scuola Superiore Sant’Anna, Italy. Cugini is one of the co-authors of the paper I will present tomorrow. It is interesting to note that Francesco's group is probably the "winner" of this year's OFC. They have 11 papers here, which is certainly the highest number for a single institution. Our paper is part of the list, since besides Cugini, Alessio Giorgetti and Prof. Castoldi are also co-authors.

They proposed an iterative algorithm to achieve global load balancing of zero-bandwidth TE LSPs (keeping it simple, these are low priority LSPs). They have realised that commercial MPLS routers present very poor balancing for this type of traffic. The algorithm is neat, and it seems to work, but I had some doubts about its scalability. I think they didn't dive into to this problem in-depth. I will probably chat with them over dinner one of these days and try to clarify this... :)

"Availability-Guaranteed Connection Provisioning with Delay Tolerance in Optical WDM Mesh Networks"

The first author and presenter was Cicek Cavdar, from Istanbul Technical Univ. (ITU), Turkey. They have exploited delay tolerance to decrease blocking in availability guaranteed shared-path-protected optical WDM networks. They reduce blocking probability without sacrificing additional resources. This is the first proposal using the time dimension to help mitigate this problem.

"LSP Request Bundling in a PCE-Based WDM Network"

The first author of this paper is Jawwad Ahmed, Royal Inst. of Technology, Sweden, but I am not sure if he was the one presenting it (I didn't pay attention in the beginning of the session). They propose the bundling of LSP requests to improve network optimization, at the expense of an increased connection setup delay. Basically these requests are sent to a centralised element (the PCE) and it may make sense to bundle these requests, to reduce the overhead in the network. Their approach showed that for a connection holding time of 20 seconds or less it makes sense doing this. But 20 seconds is a very low figure, I think, for a core network. I was not totally convinced with their results, and I think this topic is worth exploring a bit more.

"IP and Optical Integration in Dynamic Networks"

This was an invited talk by Ori A. Gerstel, from Cisco Systems (I attended a short course by the speaker on Sunday, but the talk was more interesting). The main question addressed was: "how to have a fully dynamic optical network?". He discussed the real-world needs for such dynamism, and how to enable IP/optical capabilities that will make this technology deployable. Some notes:

1. He thinks this will be a stepwise evolution (not a revolution). Some of the steps he mentioned and that interest me more are a) the creation of an automated control plane inside the optical network, b) the possibility of triggered lightpaths setup from the router, and c) optical packet switching. About this last point he is very sceptical. Me too, at least for the short-medium term: where are optical buffers and packet header processing at the optical level?

2. Some papers worth reading: "Handling IP traffic surges...", by Pongpaibool et al., and "Towards deployment of signalling based approaches...", by Salvatory e al. (ECOC 2008).

"Optical Communication Challenges for a Future Internet Design"

Another invited talk, this time a more "political" one. The speaker was Darleen Fisher, from the Natl. Science Foundation (NSF), USA. She talked about the challenges faced by the optical research community. She mainly talked about FIND (Future Internet Design), and their clean slate approach for network research. It was interesting to note that the NSF is attempting to join the optical networking people with the Internet people. One example is the DOCS project, where they are mixing people from the optical area with internet researchers (like Nick McKeown).

The questions in the end were quite interesting, with two optical researchers asking the speaker why optical research doesn't get more money, and specifically why it gets way less money than wireless projects... :)

"Survivable Logical Topology Design for Distributed Computing in WDM Networks"

Xiang Yu, from SUNY Buffalo, USA, presented an MIP formulation and an efficient heuristic for the problem of logical topology design for a distributed computing application to survive one computing cluster failure and one fibre link failure in WDM networks. This is a NP-hard problem, so he presented a MIP formulation (not scalable) and then also a heuristic. Unfortunately, he didn't manage his time well, so he skipped the most important part - the results. The little time he used for this was to say the obvious, so in the end it was not at all useful. This was the example of a talk that, for not being well prepared, wasn't able to pass a message that could be of interest to some of the audience.

"Experimental Demonstration of SIP and P2P Hybrid Architectures for Consumer Grids on OBS Testbed"

Lei Liu from the Key Lab of Optical Communication and Lightwave Technologies(OCLT), Ministry of Education, Beijing Univ. of Posts and Telecommunications, presented a paper where he proposed three types of SIP and P2P hybrid architectures for consumer grids. The idea of a hybrid architecture was interesting, turning the centralised client/server model of SIP into a P2P one. Unfortunately, again, he spent all his time explaining the architectures, and skipped the results section...

"GENI: Overview and Plans"

This was an invited presentation where, again, one could see the attempt to join the optical community with the computer nets community. GENI, the Global Environment for Network Innovations, is a suite of experimental network research infrastructure being planned and prototyped in work sponsored by the NSF, and was presented by Kristin
Rauschenbach, from BBN Technologies. The idea is that researches can use a "slice" of the infrastructure for long running, realistic experiments, so simultaneous experiments, radically different, will be running in parallel. They aim to glue together heterogeneous infrastructure (avoiding technology "lock in").

On the optics side, the GENI goal is to enable computer people direct access to optics (by virtualisation of optical capabilities and programmable optics). Someone asked what the cost for a singular researcher using GENI is, and the speaker answered that the cost would probably be part of the NSF funding contracts. So if a researcher wants to use a "slice" as part of his or her work, they do not seem to have a "cost table" for that. At least not now.

"Light-Mesh: An Evolutionary Approach to Optical Packet Transport in Access Networks"

Ashwin Gumaste, from the Indian Inst. of Technology at Bombay, India, has presented a paper proposing an alternative solution to PONs. PONs require lots of optical fibres (hence, high costs), and so their system is a "light mesh" concept, with the objective of reducing the number of fibre links needed. They have built a test bed, and they were able to achieve optical packet transport using mature off-the-shelf components. Though their system is prone to collisions, with the algorithms they use (CSMA-CD-like), they seem to achieve low delays.

OFC 2009, Day 1

2009-03-23T04:15:00.004Z

This Sunday the conference program was only some short, 3-hour long, courses, and some workshops.

The short courses are very interesting, but they have a cost, which is always a problem when you're a PhD student... but there are advantanges of being a student: for those courses that are not full, students receive an e-mail asking if they want to participate, free of charge. :) That was what happened this time, and I've enrolled to 3 short courses:

SC216 An Introduction to Optical Network Design and Planning, Jane M. Simmons; Monarch Network Architects, USA

This was a very interesting course. It started by explaining the most important optical network elements, and then it moved to algorithms - for routing, regeneration, and wavelength assignment - which was very interesting and I was not expecting in an OFC course. Then it also talked about traffic grooming. With all these things mixed, there is really a great need of good algorithms and protocols to make the network as efficient as possible. Some notes:

1. Long haul networks use roughly 80 lambdas per fibre, and regional/metro-core use roughly 40 lambdas. These are US figures, I guess, but I was impressed with these high values. Some of the networks I am aware of (from network providers) have only something between 10 and 40 Gbps, which means probably use only 1 to 4 lambdas... I guess maybe these are normal values for walled networks from a single network provider alone.

2. Today there is no wavelength conversion on optical add drop multiplexers. Wavelength conversion is only possible on OEO (Optical - Electrical - Optical) architectures. So it makes all sense the assumption we've used in the paper I will present here. We assumed there was no wavelength conversion in our all optical network scenario...

3. The optical reach (distance an optical signal can travel before necessitating regeneration) of legacy networks is between 500-600km. Today we can have 2000-4000km.

4. Having or not regeneration has immense cost implications in a long haul network. For this reason, the best routing mechanism to find the best path between two nodes should: a) from all available paths, check the one that has fewer number of regenerations, and b) then check for the shortest path in terms of number of hops.

5. Dynamic path routing may not be the best option in a long haul network - it can lead to too much "wavelength interference".

6. Using the noise figure instead of number of hops or distance as the metric for shortest path algorithms is probably the best option.

7. When a node performs regeneration, it is possible to have wavelength conversion in that node (because there is OEO conversion in that node, so we can make use of that).

8. Routing and Wavelength Assignment: doing it in separate steps (run shortest path, check if we have wavelengths available that satisfy the wavelength continuity constraint) is probably better than doing all in a single step.

9. Grooming (bundling of sub-wavelength traffic to form well-packed wavelengths) switches should be deployed in the edge, not core. And we probably don't need all nodes to be grooming-sites (maybe 20 to 40% is enough).

10. In a metro network you usually have a lot of fibres, so it's different from a long haul one, where you have less fibres but several lambdas.

SC243 Next Generation Transport Networks: The Evolution from Circuits to Packets, Ori A. Gerstel; Cisco Systems, USA

This was the course I was more interested in, but unfortunately it was the less interesting. There were two main reasons, I think, for that. The first was that most of the subject was not new to me, and since it was all just a superficial touch in the matter, I learned nothing new. The second was that the speaker wasn't particularly inspired, and the way he approached the subject was just not interesting (although one could see he really was an expert on the issue...).

Mainly Ori Gerstel talked about the two types of multiplexing used in transport networks: time division (SDH and related), wavelength (WDM) and packet (MPLS). He talked about the possible paths of evolution of all these technologies, and how they are somehow being put closer together. I was hoping he would talk more in-depth of MPLS and GMPLS, but he didn't say much rather than the very generic stuff.

There was, however, a very interesting point. He mentioned a study that looked at the household needs in 2010, in the US. In this study they consider these services would be offered to every household soon: HDTV, SDTV, PVRs, and VoIP. And they concluded that twenty such homes would generate more traffic than what traveled the entire Internet backbone in 1995! And he finished this slide using the same words I have recently used in my talks: "the bottleneck is moving to the aggregation and core parts of the network".

SC114 Passive Optical Networks (PONs), Paul Shumate; IEEE Lasers & Electro-Optics Society, USA

This was an excellent course by Paul Shumate. The objective of the course was to explain why PONs are becoming the key network approach to deliver Fiber to the Home (FTTH). Some notes:

1. The normal splitting ratio in current PONs is between 8 and 32. Although in the US they are satisfied with these values, n Europe there is a growing interest in much larger ratios, up to 2048 (European SuperPON).

2. The US seems more interested in this technology than Europe. The main reason is that in Europe the copper distances are shorter, and the copper is better (due to the reconstruction after World War II), so VDSL is seen as a good option.

3. The speaker showed the nice picture from Claffy et al. paper "the nature of the beast: recent traffic measurements from an Internet backbone", where the authors got to the conclusion that most packets in the Internet where 1500 bytes-long or less.

4. FTTH status: a) 4M homes already connected in the US (2008); b) highly successful in Asia, with 3M new subscriber each year in Japan (40-50M users in 2010/2011, probably); c) growing interest in Europe, especially Scandinavia, but also France, UK, Spain, etc.

5. How much bandwidth is necessary to the home? A: In the near future, with HDTV, many channels, TiVo, etc., at least 50-100Mbps.

6. Studies are now sugesting that symmetry is needed (asymmetric connections will be a thing of the past soon).

7. Average broadband speeds: Japan = 60 Mbps, Korea = 45 Mbps, Finland = 20 Mbps (the top three). Further below the US = 5Mbps, and the UK = 3Mbps. Portugal is in between these, with 8Mbps.

OFC 2009

2009-03-23T03:58:00.005Z

I arrived Saturday to San Diego, CA, to attend the Optical Fiber Communications conference. This Wednesday I will present here the paper "Power Excursion Aware Routing in GMPLS-Based WSONs".

I will take some notes and write a short report everyday, so if you are interested in what's hot on optical networking and optical communications, I invite you to watch this space during the next few days.

PS: To add more appetite for the topic, you can check the latest news about BT's Optic Broadband Rollout Plans.

"Mr Watson, come here. I want to see you."...

2009-03-12T07:13:00.001Z

…happened already 133 years ago:

http://www.itpro.co.uk/610140/today-in-history-the-first-telephone-call

Definition of Science

2009-03-04T12:44:00.000Z

"Science is the pursuit of knowledge and understanding of the natural and social world following a systematic methodology based on evidence."

The Science Council

The importance of stupidity in scientific research

2009-02-23T13:18:00.001Z

An excellent essay by Martin A. Schwartz. Definitely true.

More on multicast

2009-02-20T15:15:00.003Z

A very nice page on multicast:

http://www.hep.ucl.ac.uk/~ytl/multi-cast/index.html

and also a cool presentation here.

Multicast Status Page

2009-02-19T15:54:00.001Z

A nice webpage for those interested in following the growth and general status of the Multicast Enabled Internet:

http://www.multicasttech.com/status/

Measuring Net Neutrality

2009-01-29T07:44:00.001Z

Google, partners release net neutrality tools

Google and partners unveil a set of Internet performance measurement tools.

Google and a group of partners have released a set of tools designed to help broadband customers and researchers measure performance of Internet connections.

The set of tools, at MeasurementLab.net, includes a network diagnostic tool, a network path diagnostic tool and a tool to measure whether the user's broadband provider is slowing BitTorrent peer-to-peer (P-to-P) traffic. Coming soon to the M-Lab applications is a tool to determine whether a broadband provider is giving some traffic a lower priority than other traffic, and a tool to determine whether a provider is degrading certain users or applications.

"Transparency is our goal," said Vint Cerf, chief Internet evangelist at Google and a co-developer of TCP/IP. "Our intent is to make more [information] visible for all who are interested in the way the network is functioning at all layers."

In Good Gear Guide

Linear Programming

2009-01-28T11:37:00.004Z

For those that are working on optimisation of some sort, Linear Programming (LP) is a mathematical method generally used to get optimal solutions for one problem. We have an objective function that we want to maximise or minimise (maximising profits, minimising congestion), we have a series of constraints (a certain amount of budget, a certain capacity in a line), and LP gives the optimal result(s), if there is one (how much to produce to maximise profit? where to route traffic to minimise congestion?).

In my work on Traffic Engineering I bang myself quite often with LP. Since this is something new to me it is sometimes hard to grasp papers that use this technique. And I've realised that most of my colleague researchers tend not to like very much of papers that have lots of mathematical formulae... :) but sometimes they are indeed fundamental. Now I've found an excellent tutorial that explains in an amazingly intelligible and simple manner this mathematical method. Check it here:

LP Tutorial

Communications of the ACM

2009-01-27T11:06:00.000Z

Besides other interesting stuff, there is a nice discussion on network neutrality this month in Communications of the ACM (between Barbara van Schewick and David Farber). Also a nice article by Akamai’s Tom Leighton on improving the performance of the Internet (is the bottleneck moving from the last mile to the middle mile?). For those interested, check it here:

http://mags.acm.org/communications/200902/?CFID=423293&CFTOKEN=63848239

Bluesci, issue 14

2009-01-23T14:49:00.001Z

The new issue of Bluesci is now available online (pdf version).

For those who don’t know, Bluesci is Cambridge's science magazine, produced by students of the University. I have a short article on electronic paper that you can find also in the website, here.

Charles Darwin on BBC

2009-01-07T08:15:00.000Z

For those, as me, admirers of the work of the great scientist Charles Darwin, BBC is airing a season of TV and radio programmes of great interest, 200 years after his birth and 150 after “The Origin of Species”. Check everything in here:

http://www.bbc.co.uk/darwin/

I also think it is worth going to the National History Museum one of these days:

http://www.darwin200.org/

Flowers in Science

2009-01-06T22:53:00.002Z

My friend Mathias is one of the two first authors of a paper in this week's Science magazine. Congratulations, Mathias!

Bright Shiny Flowers

The study of flower color has primarily focused on chemical pigmentation. Whitney et al. (p. 130) explore how the structural features on the petal surface of a tulip flower generate color independently of chemical pigmentation via iridescence. Floral iridescence may result in an ultraviolet signal that is visible to insects. Bumblebees can learn to use information from iridescence to select among potential food sources. Thus, iridescence may contribute to plant-pollinator interactions.

Abstract: http://www.sciencemag.org/cgi/content/short/323/5910/130

Full Text: http://www.sciencemag.org/cgi/content/full/323/5910/130

Intel breaks record with optical CMOS device

2008-12-10T09:39:00.000Z

Silicon-based photo detector beats III-V components

SAN JOSE, Calif. — Researchers from Intel Corp. have demonstrated a photo detector built in CMOS that the company claims is the highest performance optical component of its class to date. The avalanche photodetector (APD) described in a paper in the journal Nature Photonics shows the way to designs that could increase the distance or lower power and cost of optical links, Intel said.

The research effort is one of many small steps forward in silicon photonics in recent years from Intel. The company aims to commercialize some of its work in PC platforms in as little as two to three years, said Mario Paniccia, director of Intel's photonics lab who reported the advance.

Intel's APD achieved a gain-bandwidth product of 340 GHz, higher than any previous device made in any process technology. The metric is a broad measure of the component's signal amplification capability at any given speed.

"This is the first time a silicon photonics device has a better performance than a III-V device, in this case specifically indium phosphide," said Paniccia. "We started with goal of getting [in silicon] 90 percent of the performance of [more] exotic materials with an order of magnitude less cost, but we now have a silicon devoice that is better performance than anything measured in indium phosphide," he added.

APDs are primarily used today in relatively costly modules enabling optical links at 10 Gbits/second over tens of kilometers. The Intel APD could support devices with throughput up to 40 Gbits/s at an order of magnitude less cost, Intel said.

In EE Times

Abstract of the Nature Photonic paper in here, full paper in here

Ant routing and congestion control

2008-12-07T17:38:00.001Z

A very interesting paper:

Analytical and Numerical Investigation of Ant Behavior Under Crowded Conditions

Maybe we can build better routing protocols by learning from ants... :)

Status of IPv6 deployment

2008-12-07T17:36:00.001Z

A nice presentation on the status of IPv6 deployment. Quite informative:

Measuring IPv6 Deployment

Andrew Odlyzko talk

2008-11-27T09:34:00.004Z

Yesterday I had the privilege to attend a talk by Andrew Odlyzko. It was definitely one of the best talks I've ever been to. The title was "Internet evolution and misleading network paths".

He started by showing a very interesting graph that illustrates that research is usually out of sync with reality:

The graph shows the number of publications per year (from First Monday). As one can see, the 1990s seemed to have been the ATM research decade, while Ethernet was lagging way behind... we know where we are now, so Andrew made his point very clear.

He continued by presenting several misleading dogmas that have impeded reform and restructuring of networks (one example: "carriers can develop new services"), and the rest of the talk focused on these dogmas.

His main claim was that the threat to the Internet was not much traffic. On the contrary, it is little traffic. He argues by showing results from the MINTS project in which he is involved: traffic growth rate is slowing down, and this figure is now of 50%/60% per year, globally (but with exceptions). He made some simple calculations to show that with this growth there is no need to increase expenditure on networking: if we increase traffic from 100 to 150, and we assume a decrease of unit cost from 100 to 67 (33% seems an acceptable number), then total cost will be the same... But, he underlined, if broadcast TV moves itself to the Internet (especially with HDTV), things will surely change. Abruptly.

He also mentioned cloud computing: "don't expect everything in the cloud soon - or ever!". He argues that there is too much information in computers, and that would take literally years to put everything from your disks to the cloud (transmission is lagging way behind). But one of my PhD supervisors, Jon, criticises this (correctly, in my point of view) in one of his blogs (a great summary of the talk, by the way): we only need to transmit the residual.

Another interesting point was the idea that volume is not value. Again, absolutely correct, and excelent arguments. Check the revenue per MB of each of these services: SMS = $1000; Cellular call = $1; Fixed call = $0.1; Residential net = $0.01; backbone traffic = $0.0001.

He also claims, with reason, that the long haul is cheap, compared to the rest: "long haul is not where the action is".

In the past, as today, the dominant types of communications are business and social, not content, he added. And he stressed the importance of oral communication with a superb example: "would we attend a talk if there were no slides?". Everybody's answer: "yes". And would we attend if no one was allowed to speak (including the speaker)? Everybody's answer: "no". Good point! "And why don't we have a non-disruptive (like e-mail) voice messaging service yet?"

He also pointed out that there is no point in streaming real data - we need faster than real time (Jon has mentioned this too very well in his blog mentioned above).

He finalised with predictions for the future network. I close this post with them:

1. dumb pipes

2. over provisioning ("waste that which is plentiful", George Gilder)

3. dominated by computer to computer interaction, driven by human impatience

4. horizontal layering, structural separation

5. market segmented by size of dumb pipe

Video Is Dominating Internet Traffic, Pushing Prices Up

2008-11-03T11:49:00.000Z

You are watching a lot more video on the Internet, and you may start to pay your Internet provider more for it. That was one of the conclusions I walked away with after spending Friday at the annual seminar of the Columbia Institute for Tele-Information at the business school of Columbia University. The theme was “The Dawning of the Ultra-Broadband Era,” and it explored the implications of Internet service that is far faster than the current offerings over phone wires, cable and wireless connections.

in NYTImes

University researchers developing cancer-fighting beer

2008-10-24T07:04:00.001+01:00

Genetic engineering could give Joe Six Pack anti-aging and cancer-fighting benefits

October 21, 2008 (Computerworld) Have you ever picked up a cold, frosty beer on a hot summer's day and thought that it simply couldn't get any better?

Well, you may have to think again.

A team of researchers at Rice University in Houston is working to create a beer that could fight cancer and heart disease. Taylor Stevenson, a member of the six-student research team and a junior at Rice, said the team is using genetic engineering to create a beer that includes resveratrol, the disease-fighting chemical that's been found in red wine.

In Computer World

Comment: Now this is clearly a good example of great research: you mix the finest ingredients, red wine and beer, and you get something approximate to the elixir of life. (Not that we needed another excuse to drink it… J)

"Making Internet routing scale with shim6"

2008-10-06T18:57:00.003+01:00

Iljitsch van Beijnum, a PhD student at the Universidad Carlos III de Madrid, gave a very interesting talk last week here in the lab. Iljitsch has contributed to the IETF Multihoming in IPv6 working group, and has already written two books: BGP: Building Reliable Networks with the Border Gateway Protocol, and Running IPv6.

The talk was about multihoming (one user using multiple ISPs - that's an easy way to put it) and Shim6, as an alternative technique to solve scalability issues in BGP related to multihoming and IPv6.

It is a fact that the routing system will not be able to absorb millions of multihomers. The IETF has been dealing with this since 2001, and the outcome was Shim6. With Shim6 each user has one address per ISP, but all these addresses are hidden from the upper layers (these are inserted between IP and TCP/UDP, hence the name "shim"). This then allows hosts to move ongoing communications from one set of addresses to another.

Most fundamental problems I've perceived in Shim6: 1) only works with IPv6; 2) both sender and receiver must "understand" Shim6 - if one of them doesn't understand they'll just communicate as usual.

Iljitsch also talked about his current work on multi-path TCP. A nice thing about his work is that he his trying to build mTCP with changes only in the sender - that way the receiver is oblivious of the use of a "new" technique.

KDDI to launch 1Gbps fiber-optic service in Oct

2008-09-29T14:13:00.001+01:00

KDDI Corp will launch a fiber-optic communications service with upload and download speeds each of up to one gigabit per second on Oct 1. The new service will target people living in single-family homes and low-rise apartment buildings. The traffic speeds will be the fastest in eastern Japan, up drastically from the current 100 megabits per second.

in Japan Today

Major EU P2P research project hopes to kill traditional TV

2008-07-21T11:26:00.000+01:00

Dutch academic Dr. Johan Pouwelse knows BitTorrent well, having spent a year of his life examining its inner workings. Now, as the scientific director of the EU-funded P2P-Next team, Pouwelse and his researchers have been entrusted with €19 million from the EU and various partners, and what they want in return is nothing less than a "4th-generation" peer-to-peer system that will one day be tasked with replacing over-the-air television broadcasts.

in ars technica

IP traffic to 'double' every two years

2008-06-19T10:14:00.000+01:00

The report by Cisco (link below) is short and has some interesting figures - even if you tend to disbelieve in these forecasts, there are some nice figures from 2007 (for example although p2p file sharing traffic grew its percentage decreased from 60% to 51% due to the increase in video traffic and also an increase of web-based file sharing in some regions).

Lots of work for network researchers ahead... ;)

IP traffic to 'double' every two years

Web traffic volumes will almost double every two years from 2007 to 2012, driven by video and web 2.0 applications, according to a report from Cisco Systems.. Increased use of video and social networking has created what Cisco calls 'visual networking', which is raising traffic volumes at a compound annual growth rate of 46 per cent.

Cisco's Visual Networking Index (PDF) predicts that visual networking will account for 90 per cent of the traffic coursing through the world's IP networks by 2012.

The upward trend is not only driven by consumer demand for YouTube clips and IPTV, according to the report, as business use of video conferencing will grow at 35 per cent CAGR over the same period.

Cisco reckons that traffic volumes will be measured in exabytes (one billion gigabytes) by 2012 and will reach 552 exabytes by that time.

in iTnews

Cheap wii

2008-06-18T12:27:00.001+01:00

If this works as it looks it's a really cool and cheap wii. :)

http://www.camspace.com/

Nokia Morph Concept

2008-06-18T12:26:00.000+01:00

Quite cool...

http://www.youtube.com/watch?v=IX-gTobCJHs&eurl

What a Botnet Looks Like

2008-05-09T11:20:00.001+01:00

A really nice tool to check the inners of a botnet.

Network Neutrality is Dead

2008-05-06T11:22:00.003+01:00

So UK ISPs are furious with BBC's iPlayer and want the broadcasting company to fund their network's upgrade. In the meanwhile in the US Comcast is blocking BitTorrent traffic. Network neutrality ("a load of bollocks", as Virgin Media's CEO puts it) had better days...

It's getting congested out there! :)

Telco IPTV Households Now 14 Million Worldwide, says Parks Associates

2008-04-16T12:23:00.001+01:00

The number of households worldwide using telco/IPTV services grew from 4.7 million in 2006 to exceed 14 million in 2007, according to a research whitepaper by Parks Associates. Europe leads with nearly 60 percent of telco/IPTV households worldwide. The analysts define Telco/IPTV as landline-based multichannel and on-demand video services provided by a telephone operator. Free whitepaper.

in Tekrati

Interdomain Multicast Solutions Guide

2008-04-16T09:52:00.000+01:00

Maybe an interesting book (the best thing is that it's free):

http://ccnacertificate.blogspot.com/2008/04/interdomain-multicast-solutions-guide.html

France dominates European IPTV

2008-03-12T08:44:00.001Z

Country has 75 per cent of IPTV subscriptions in Western Europe

The latest World Broadband Information Service from Informa Telecoms & Media noted that France has nearly 10 times as many IPTV subscriptions as second-placed Spain and close to 75 per cent of the Western Europe total.

in vnunet.com

NOSSDAV 2007

2008-01-31T12:03:00.000Z

NOSSDAV 2007 has some rather interesting papers on IPTV, so it is worth to check the conference's website.

Internship in Telefonica R&D, Barcelona

2008-01-31T11:54:00.000Z

I have been a little quiet for some time because I am now in Telefonica Research Labs in Barcelona, so I've been quite busy moving, finding a place, etc. I am here for a 3 or 6 month internship as part of my PhD. I am working on "user profiling for IPTV". I have lots (hundreds of GB!) of interesting data from Imagenio (Telefonica's IPTV service) - switching logs, EPG (Electronic Program Guide) information - to analyse.

This data is very useful for user profiling and recommendation systems (to be used for targeted advertisement, for instance), and I am sure will also be useful for my PhD research, as input to my simulations.

IPTV subscribers worldwide forecast to reach 38.4m by 2012

2007-12-13T07:59:00.000Z

A new report that there will be a huge growth in the number of IPTV subscribers worldwide by 2012, increasing seven-fold from the end-2006 figure to reach 38.2 million. Revenues for service providers will possibly rise ten-fold.

These is only a prediction, and one can believe on it or not, but the fact is that service providers today seem to believe that IPTV will be the next killer application. A paper written last month, in IEEE Communications Magazine, was thus properly entitled "Internet Protocol Television (IPTV): The Killer Application for the Next-Generation Internet". Although this paper is not very well written and makes some confusing and sometimes even erroneous remarks (like confusing DSL filters cutoff frequency... twice!) - which is strange in one such a high level magazine - it is nevertheless worth reading.

KOREA: LG Dacom launches IPTV service

2007-12-13T07:47:00.000Z

LG Dacom, Korea's third-largest fixed-line operator, yesterday launched the 3rd IPTV service in that country. While most contries worldwide are only now offering their first IPTV, South Korea has now 3 companies offering the service, which really illustrates their almost leading position in broadband services.

Central America's First IPTV Service

2007-12-07T10:32:00.000Z

Panama’s CTV Telecom is the first operator offering IPTV service in Central America. It has selected the Amino AmiNET125 multi-codec Set-Top-Box (STB) for its service.

CTV will build a 100% fibre optical network in Panama City (the capital) exclusively to deliver the new IPTV service.

BCS Roger Needham Lecture

2007-11-28T16:03:00.000Z

One of the nicest things of being a student in Cambridge is the incredible number of excelent talks, seminars, workshops, etc., available. It is very difficult to choose what events to go to each week.

Last week I went to London to attend a lecture by Professor Mark Handley. And what a great lecture it was! Professor Handley has done very significant research on computer networks for the past decade or more, and he is the author of several Internet standards such as SIP and PIM-SM (the de facto IP multicast protocol used by nowadays’ service providers). He is also the co-author of two books, both written together with one of my PhD supervisors, Professor Jon Crowcroft. Here are links for both books: The World Wide Web - Beneath the Surf and Internetworking Multimedia. Professor Handley was this year's winner of the Roger Needham award.

The topic – “Evolving the Internet: Challenges, Opportunities and Consequences” – was very interesting, and Mark touched very sensible points. As a very general summary, the main idea he passed was that the evolution of the Internet must be made with the use of flexible technology. He pointed some problems that should be addressed in the actual Internet: its openness (leading to atacks), congestion problems, and BGP (the de facto standard for interdomain routing). He also gave nice examples taken from his own experience: IP mcast (not totally successful; it is not global as they wanted, it is only used inside single ISPs) and SIP (very successful).

Imagenio grows and grows...

2007-11-21T22:52:00.000Z

Telefonica's Imagenio IPTV service has 469,000 subscribers already (12% share of the Spanish pay-TV market). And they are now also on the Czech Republic with 53,000 customers.

2007-11-20T09:43:00.000Z

Two very interesting breakthroughs in manipulation and slowing down light are reported in AlphaGalileo, one from the University of Bath and the other from the University of Surrey. We might be near the so-expected "optical capacitor", which could be a giant leap for optical fibre comms.

10 MPLS traffic engineering myths and half truths

2007-11-07T22:50:00.001Z

Ivan Pepelnjak talks about some myths and half truths concerning MPLS-TE. Some are quite obvious, but nevertheless it is worth the look.

Internet Research Needs Better Models

2007-10-30T11:48:00.000Z

Another interesting paper I've discovered these days - published in 2002 - was Internet Research Needs Better Models, by Sally Floyd and Eddie Kohler (Berkeley). In there they conclude that Internet research has a great need for better models, and for better common evaluation of these models. Some of the wrong implicit assumptions that they point down as the cause of the problem are: 1) flows live for a long time and transfer a lot of data, 2) "dumbbell" topologies with one congested link are used, 3) flows on the congested link share a small range of round-trip times, 4) most data traffic across the link is one-way, 5) reverse-path traffic is rarely congested.

In the exact moment that I am specifying in detail my PhD work plan (thinking of source, demand and topology models, evaluation, etc.) it is really important to take the above-mentioned points into consideration.

Internet Measurement Conference 2007

2007-10-29T17:32:00.000Z

The program of the Internet Measurement Conference 2007 was very interesting. The best paper award went to I Tube, You Tube, Everybody Tubes: Analyzing the World's Largest User Generated Content Video System. Mia (that has been around in the Computer Lab for some time), Pablo Rodriguez (from Telefonica) et al presented a very detailed and interesting study on User Generated Content systems, like youtube.

I also enjoyed the article Characterizing Residential Broadband Networks. Some of the results were interesting: 1) packet transmissions over cable suffer high jitter as a result of cable’s time-slotted access policy (which can affect TCP Vegas due to its reliance on RTT), 2) DSL links show large last-hop delays and considerable deployment of active queue management policies such as random early detection (RED), 3) both cable and DSL ISPs use traffic shaping and deploy massive queues that can delay packets for several hundred milliseconds (really bad for things like VoIP).

It was also interesting to know that about 83% of IP packets are all-zero in the Type of Service field, and IP options are virtually not used, along with fragmentation. Oh, and Path MTU discovery rules... :) (check more on Analysis of Internet Backbone Traffic and Header Anomalies observed)

But the program is interesting as a whole and hence worth inspecting in detail.

Multicast TE still remains in its preliminary stage

2007-10-18T17:49:00.000+01:00

I have just found out a very interesting site on QoS Aware Multicast Management. In there they state that "relevant research on multicast TE still remains in its preliminary stage compared to its unicast counterpart".In fact, "despite the progress for unicast services, traffic engineering for multicast flows remains largely a dark area till now".

They have a specific section on Multicast Traffic Engineering. In there they divide Multicast TE in MPLS Based Multicast Traffic Engineering and IP Based Multicast Traffic Engineering.

In MPLS-based multicast TE point-to-multipoint LSPs are explicitly constructed, and the mathematical problem formulation is often related to the Steiner tree problem for minimising traffic delivery cost. So we need to set up dedicated MPLS tunnels. My PhD proposal is exactly on this issue, and part of it is to develop scalable heuristics for multicast TE.

In IP Based Multicast Traffic Engineering, the basic idea is to convert engineered multicast trees into shortest path trees by optimising link weights. The main advantage in this approach is that legacy IP routers are able to compute multicast trees by applying the Dijkstra’s shortest path algorithm (with the pre-configured link weights).

A Machine Learning Approach for Efficient Traffic Classification

2007-10-18T17:27:00.000+01:00

Today's talk on the Computer Lab was on "A Machine Learning Approach for Efficient Traffic Classification". The speaker was Wei Li. Wei Li was a PhD student in QMUL but now has moved to the Computer Lab "following [his] supervisor Andrew Moore".

The talk was on the very interesting subject of online traffic classification. Besides being used for network monitoring and intrusion detection, traffic classification can also serve as the input for application modeling. For that reason it could be useful to model the IPTV service, for instance, which would in turn be important as an input to my OPNET simulations.

Wei Li presented a machine-learning approach that accurately classifies internet traffic using C4.5 decision tree. Without inspecting packet payload, their method can identify traffic of different types of applications with 99.8% total accuracy, by collecting 12 features at the start of the flows. But overall it was a very nice talk.

Some questions arose on the features chosen (why those 12 and not others, when 248 were available?) and on the relationship between accuracy and time (can be an important aspect to consider in some applications, like intrusion detection).

How to read a paper

2007-10-09T16:16:00.000+01:00

A big part of a researcher's time is spent reading papers. Either with the aim of keeping current in their field or for a literature survey in a new field, a researcher will read tens (or even tons :) ) of papers monthly. For this reason, it is very important to know how to read a paper.

In July's issue of Computer Communications Review (CCR Volume 37, Number 3 July 2007) S. Keshav (he is the author of a very nice book on communications systems: An Engineering Approach to Computer Networking) wrote a good article on "how to read a paper". I believe a perusal of this short paper is crucial for a new researcher. Paper reading skills are not taught anywhere, so this article is extremely useful.

An Introduction

2007-10-03T15:14:00.000+01:00

I am a PhD student in the University of Cambridge. I am now starting my second year, and I would like to share all kinds of stuff related with my research work. This was the reason why I've decided to create this blog: to be sort of a logbook of my research.

The objective of my PhD thesis is the design of a GMPLS system for provisioning optical core and access networks for Multicast TV, and so this weblog will be dedicated to topics related - although not exclusively. As perceived by the blog title, I consider all of this "IPTV research" generically. I am pretty sure these topics will eventually be addressed:

1) Multicast
2) Traffic Engineering (TE)
3) Optical Networks
4) Routing
5) Quality of Service
6) IPTV
7) ...

and mixes of all the above (something like "Multicast TE Routing in Optical Networks to provide IPTV with nice QoS" would be a cool topic :)). And more generic topics related to telecommunications, Internet, protocols, next generation networks, etc., etc., will probably appear too.

I will try to bring news, scientific papers (old and new), ideas, results, etc. I will also welcome comments from all of my readers, either directly posted into the site or to my e-mail: fvramos at gmail dot com.