Each summer the Systems and Networking Research Group at Microsoft Research has several outstanding students from top schools join the group for research internships. Many of the research projects they have pursued have resulted in refereed research publications, including those winning best paper awards.

This page lists the interns who have worked with us in 2002.

2002 Research Interns

Ranveer Chandra, Cornell University

This summer I worked on different wireless projects with Victor and Lili. I built a command line utility, called WRAPI, to query and set various parameters of a wireless connection. I also worked on enabling simultaneous connections to multiple networks using a single wireless card. Victor and I built this system and demonstrated its feasibility. Finally, Lili, Kamal, Mohammad and I worked on developing good algorithms for ITAP placement in a mesh network.

Kyle Jamieson, Massachusetts Institute of Technology LCS

This summer I worked on an ad hoc networking project. We looked at different types of radios and medium access control protocols, as well as ad hoc routing and higher-layer issues. We also investigated the physical layer, looking at directional antennas and other types of hardware.

I also worked on protocols for base-station environments where there is a second, lower data rate channel. The second channel can be used for control packets, and I investigated how such a system could improve the performance of TCP over schemes that don't use control channels. This work was done in the context of power-savings for the mobile device.

Karthik Lakshminarayanan, University of California, Berkeley

P2P research makes a lot of assumptions about network characteristics experienced by the end-host. One of the most common things is assuming correlation between bandwidth and delays. While some bit of measurement and analysis has been done to this end in networking research, it is still unclear how the broadband hosts behave. It is probably fair to believe that broadband hosts would form a bulk of the future p2p client set. In the same vein, it would also be interesting to study the temporal variation of characteristics such as delays between broadband hosts.

To this end, along with my mentor, Venkat Padmanabhan, I worked over this summer to develop a software dubbed "PeerMetric" which would be installed on broadband hosts. This would then perform periodic network measurements with other clients and report the results back to the server. By collecting such measurements from a diverse set of end-hosts, we plan to conduct an extensive study of the characteristics of broadband hosts.

Patrick Reynolds, Duke University

I worked on Farsite, a scalable distributed filesystem in which replication and Byzantine protocols are used to provide high availability storage using mutually untrusting client workstations. My portion of the project was implementing SALAD, a low-diameter peer-to-peer overlay, and using it to find duplicate files in Farsite clusters. Coalescing duplicate files is important because it can double the amount of space that each user can use without reducing the number of replicas of each file stored. I have also worked on improving SALAD's performance and robustness, in hopes that it can be used for other applications outside of Farsite.

Stefan Saroiu, University of Washington

This summer, I have been having a lot of fun working on the Herald project with Mike, Marvin, Alec and Nick. Herald is a globally scalable distributed publish/subscribe event notification system. An event can be published anywhere in the world, and the Herald system will ensure timely and reliable event delivery to all event subscribers. The Herald servers self-organize themselves into an overlay and run a multicast protocol to distribute the events among themselves.

In the first half of the internship, I was part of the effort of demoing the first live Herald system (and not a simulation): a cluster of 16 machines running an overlay of 480 Pastry nodes with Scribe doing multicast groups in order to publish and deliver events to subscribers. For the second half, I participated in the design and implementation of a new distributed hash table (SkipNet) that addresses the controllability and data placement issues in the current P2P schemes. In its current design, SkipNet maintains probabilistic logarithmical bounds on the length of overlay paths and the amount of state per each participating peer (the same way as Pastry and Chord do). In addition, SkipNet has the ability to control the location and responsibility of hosting objects and data.

Adam Stubblefield, Rice University

This summer I worked on a authentication scheme based on random "inkblots." The system is designed to help users choose and remember strong passwords. I have designed and implemented a prototype inkblot generator and authenticator. I've also been working on validating the usefulness of the system using both evidence from psychology literature and results from small user studies.

An Zhu, Stanford University

Project: Theoretical aspects of the EDN project

This summer I've been working with Yi-Min Wang (my mentor) and Lili Qiu on Event Distribution Network project. This project is essentially concerned with a publisher/subscriber model in the distributed computing setting. As an event occurs, the system has to figure out which subscriptions does the event effect and send out notification messages to the corresponding users. The goal of the distributed system is to evenly spread the total work among machines so that to achieve overall optimal throughput performance. There are many objectives one can try to optimize in such a system, currently; we formulated a theoretical problem that captures the load balancing criteria in such a system. The problem is NP-complete and we are investigating approximation algorithms for such a problem.

Systems and Networking Research Group | Interns and visitors from all years