Paper Presentation

• Please check that you have a slot in the schedule (linked below).


• You have to choose a paper at least 10 days before your presentation.
  Email me the name of the paper, and a URL that I can link to.
  If no electronic version is available, then you should make copies for all the students (we can provide you the copier).


• How to choose a paper?
  Look in Possible Papers for Presentation.
  Or look in recent conference proceedings (SPAA, PODC, ICDCS, SIGCOMM, INFOCOM).


• You should meet me at least once to discuss the contents of the paper.


• Optionally, you can also meet me to discuss the slides and the presentation itself.


• Guidelines for the presentation:
  • You have 30 minutes to speak + 5 minutes for questions/discussion.
  • You should make slides, either transparencies or electronic.
  • The thirty minutes can be roughly divided as follows:
    • 10 minutes for the Introduction and Motivation (why is the problem important?.)
    • 15-20 minutes for the body of the paper
    • 5 minutes for the conclusions.

Presentation Schedule

This course studies the new algorithms behind the current generation of internet applications.

• Algorithms for Web Search

  Current search engines have the capacity to rank the results of web searching according to importance and very often, their ranking matches our own intuitive ranking of these pages. How?

  • Kleinberg's algorithm
  • Google's PageRank Algorithm

  Readings:

  • Authoritative Sources in a Hyperlinked Environment by Jon Kleinberg
  • The Anatomy of a Large-Scale Hypertextual Web Search Engine by Sergey Brin and Lawrence Page
  • The PageRank Citation Ranking: Bringing Order to the Web by Page, Brin, Motwani,and Winograd
  • Topic Sensitive Pagerank by T.Haveliwala

• Algorithms for Massive Data Sets
  • The space complexity of approximating the frequency moments by Noga Alon, Yossi Matias, Mario Szegedy
    Lecture slides for Feb 11th: One pass computation of the second frequency moment (ps)
• Peer-to-peer systems

  Peer-to-peer systems are a fundamentally different way to share resources (data, processor cycles, etc) over the internet. In the ideal peer-to-peer network, autonomous computers can join and leave the network easily, without need for global coordination, and users can locate resources quickly.

  A basic problem is distributed data location: efficiently tracking the location of an object in this changing distributed system. We will study various peer-to-peer systems and their underlying algorithms.

  • First Generation Systems - Napster, Gnutella
  • More sophisticated ones (using distributed hashing) - CAN, Chord, Tapestry

  Readings:

  • A Scalable Content-Addressable Network by Ratnasamy, Francis, Handley, Karp and Shenker
  • Chord: A Scalable Peer-to-peer Lookup Protocol for Internet Applications by Stoica, Morris,Karger, Kaashoek and Balakrishnan
  • Tapestry: An Infrastructure for Fault-tolerant Wide-area Location and Routing by Zhao, Kubiatowicz and Joseph
  • Accessing Nearby Copies of Replicated Objects in a Distributed Environment by Plaxton, Rajaraman, Richa
  • The Arrow Distributed Directory Protocol by Demmer and Herlihy
  
  Lecture slides for the Arrow protocol: (ppt)

• Web Page Caching

  Readings:

  • Web Caching with Consistent Hashing by Karger, Sherman et. al.
  • Consistent Hashing and Random Trees by Karger, Lehman et. al.
  • Bloom Filter.
    Burton Bloom, Space/time trade-offs in hash coding with allowable errors, CACM, 13(7):422-426, July 1970
  • Network Applications of Bloom Filters: A Survey by Broder and Mitzenmacher

• Other Topics (coverage depends on time and interest)

  • Scalable High Speed IP Routing Lookups by Waldvogel, Varghese, Turner, Plattner
  • The Impact of Internet Policy and Topology on Delayed Routing Convergence Labovitz, Ahuja, Venkatachary, and Wattenhofer
  • Use of ideas from Economics in Network Routing.
    The Internet is made up of many independent entities, each out to maximize its own profit. But, these entities have to coordinate in order to do any distributed task, such as routing a packet from one node to another. Can we design pricing schemes so that the entities have no motivation to cheat?

    • Algorithms for Selfish Agents by Noam Nisan
    • The powerpoint talk by Papadimitriou, linked off this page

  • Byzantine Agreement (from the text by Attiya and Welch)

Course requirements

• Do a project of your choice (theoretical or programming)
• Discuss a research paper in class
• Class participation
• Homeworks
• No exams

Office Hours: Tuesday and Thursday after class, or whenever you can find me in my office. Or, send me an email and set up an appointment.

Possible papers for presentation

Assorted Links

Course Information

• Course Number: CprE 594, Section ST, Ref. # 2992 005
• Class Location: Coover 1219
• Meeting Times: Tue and Thu, 11.00 am - 12.20 pm
• Counts for 3 credits
• There will be no textbook, and most material will be taken from recent research papers.