CSE 481G: P2P Distributed Systems Capstone

Location: Tu/Th 12 – 1:30, EEB 03

Pre-requisites: 303, 326, 341, 373, and one of 490H, 451 or 461

The last few years has seen a blizzard of practical and widely used large scale distributed services. While cluster-based services (such as Google, Akamai, and Amazon) have received most of the attention, peer to peer (p2p) systems such as BitTorrent and eDonkey have emerged as an alternate model for supporting very large scale applications. Given the rapid increases in end host computation, storage, and in some cases bandwidth capacity, the question becomes: can we build google-scale services without a google-scale infrastructure, by leveraging end hosts?

This will be a large project design course, to build some interesting software artifact in the p2p space in a sizable team. To aid the project construction, we have some initial building blocks, and lectures will cover some of the design principles underlying p2p systems, including the challenges of scale, incentives, failures, consistency, and security. The course will count as a capstone for the computer engineering software track, and so a project report and presentation will be required. Some reading of research papers will also be involved, but the primary focus will be on the project design and construction.

Example projects include (student suggested projects welcome):

Build a p2p system capable of streaming millions of live video streams, such as March Madness.

Build a p2p version of MapReduce.

Build a p2p version of Akamai.

Build a p2p replacement for DNS.

Build a p2p system for automatic backup storage.

News

Project proposal requirements.

General project advice.

Each group should have at least one member sign up for an emulab or planetlab account, here are links to the emulab tutorial, planetlab tutorial and plman download page.

The message board is up, if you have ideas for projects, post it here.

Course Schedule

Week	Tuesday	Thursday	Friday Deliverable
1	April 1: Project brainstorming; A brief history of P2P (if time)	April 3: More project brainstorming P2P tool overview Emulab; PlanetLab; PLman	April 4: Team membership and pre-proposal
2	April 8: P2P File Sharing Read: Cohen, Incentives Build Robustness in BitTorrent	April 10: P2P Game Theory Read: Piatek et al., Do Incentives Build Robustness in BitTorrent?	April 11: Project Proposal
3	April 15: No class	April 17: no class
4	April 22: P2P Workloads Breslau et al., Web Caching and Zipf-like Distributions	April 24: P2P Economics Gray, Distributed Computing Economics, Microsoft TR, 2003.	April 25: First project update
5	April 29: Project presentations and discussion	May 1: Project presentations and discussion
6	May 6: DNS Read: Mockapetris and Dunlap, The Development of the Domain Name System	May 8: Distributed Hash Tables Read: Maymounkov and Mazieres, Kademlia: A P2P Information System	May 9: Second project update
7	May 13: Project presentations and discussion	May 15: Project presentations and discussion
8	May 20: Debugging P2P Systems Read: Lamport, Time, Clocks and the Ordering of Events in a Distributed System	May 22: State Management Read: Lamport, Paxos Made Simple	May 23: First draft of project report
9	May 27: Project presentations and discussion	May 29: Project presentations and discussion
10	June 3: Project discussion	June 5: no class
Finals Week	June 10: Final project report	June 12: Project Presentations