Tomas Isdal

I am currently a Ph.D. Student at the University of Washington, working with Tom Anderson and Arvind Krishnamurthy. My interests include measurement of Internet path properties and design of peer-to-peer and swarming systems, specifically bridging the gap between traditional centralized content distribution and peer to peer based content distribution in terms of performance.

I graduated with a Masters degree in Computer Science and Engineering from The Royal Institute of Technology in Stockholm, Sweden in October 2006 after finishing the last year of the degree at the University of Washington.

You can reach me at:

Research

• BitProbes
  Traditional methods of conducting measurements to end hosts require sending unexpected packets to measurement targets. Although existing techniques can ascertain end host characteristics accurately, their use in large-scale measurement studies is hindered by the fact that unexpected traffic can trigger alarms in common intrusion detection systems, often resulting in complaints from administrators. We have developed a measurement system that works around this challenge by coordinated particapation in popular peer-to-peer BitTorrent swarms. BitProbes is able to unobtrusively measure bandwidth capacity, latency, and topology information for 500,000 end hosts per week from only eight vantage points, an order of magnitude more than previous measurement systems that map edge networks. Read more at the BitProbes homepage

• BitTyrant
  BitTyrant is a lab-created BitTorrent client, based on the findings from both the BitProbes end host measurement and the BitTorrent incentives paper. By managing uplink resources more intelligently, BitTyrant is on average 70% faster for clients with a 1 megabit uplink capacity than the currently most popular BitTorrent client. Read more on http://bittyrant.cs.washington.edu.

• "PlMan" or cPlane
  The cPlane or "PlMan" is an application designed to simplify the process of deploying and running experiements on PlanetLab.

• iPlane
  iPlane is a scalable service providing accurate predictions of Internet path performance between two arbitary IPs.

Publications:

• One hop Reputations for Peer to Peer File Sharing Workloads [ PDF]
  Michael Piatek, Tomas Isdal, Arvind Krishnamurthy, Thomas Anderson
  In the 5th USENIX Symposium on Networked Systems Design & Implementation (NSDI '08), April 2008, San Francisco, CA.

• Leveraging BitTorrent for End Host Measurements. [ PDF]
  Tomas Isdal, Michael Piatek, Arvind Krishnamurthy, Thomas Anderson
  In the 8th Passive and Active Measurement Conference (PAM '07), April 2007, Louvain-la-neuve, Belgium

• Do incentives build robustness in BitTorrent? [ PDF ]
  Michael Piatek, Tomas Isdal, Arvind Krishnamurthy, Thomas Anderson
  In the 4th USENIX Symposium on Networked Systems Design & Implementation (NSDI '07), April 2007, Cambridge, MA. Awarded best student paper

• iPlane: An Information Plane for Distributed Services. [ PDF ]
  Harsha V. Madhyastha, Tomas Isdal, Michael Piatek, Colin Dixon, Thomas Anderson, Arvind Krishnamurthy, Arun Venkataramani
  In the 7th USENIX Symposium on Operating Systems Design and Implementation (OSDI '06), November 2006, Seattle, WA

• Using BitTorrent for Measuring End-To-End Internet Path Characteristics. [ PDF ]
  Tomas Isdal
  Masters Thesis, The Royal Institute of Technology (KTH), October 2006, Stockholm, Sweden

Class Projects

• CSE 561 (Graduate Networks)
  For the class project, we implemented a multi-hop ad hoc network that used sound as the medium of communication. Using only commodity microphones and speakers we achieved a maximum transfer rate of 128 bps. Though we suffered migraines (due to the high pitched beeps which we used to encode bits), the project was a resounding success (literally!). It was joint work with John P. John and Tanya Bragin. The final report and source code is available for download.
• PageRank(ing)
  As a class assingment in CSE454 we used different techniques for ranking search engine results. I used the open source webcrawler nutch and implemented different ranking algorithms including TF/IDF for both pages and anchortext, proximity of search terms, and Pagerank as described by Sergey Brin and Lawrence Page in 1999. The Pagerank source code and the other ranker implementations can be found here