CS856: Internet Server Performance

Reading Schedule

For each paper (except "Paper Reading Check List", and "How to Read a Research Paper") you must produce a summary of the paper that adheres to the guidlines/outline below (see Summary Outline). Each Summary is limited to 3/4 page 12 point font and must include everything (i.e., your name, paper title, etc.) If two papers are being done in one day both reviews/summaries must be printed on one page (i.e., double sided).
Note: I am extremely serious about the limits and the format of these summaries. If they are not followed precisely you will get no credit for doing them.

List of possible readings (i.e., links to papers can be found here)
Summary Outline (use this)
Tuesday May 9 (Note: Summaries are only required for the Lampson paper)
- "Paper Reading Check List", Sugih Jamin, jamin@eecs.umich.edu.
- "How to Read a Research Paper", by Spencer Rugaber
- "An Evaluation of the Ninth SOSP Submissions or How (and How Not) to Write a Good Systems Paper", R. Levin, D. Redell, Operating Systems Review, July 1983, pp. 35-40.
- "Hints for Computer System Design", Butler W. Lampson, IEEE Software, 1(1):11-28, January 1984. (also appears in Proc. of the 9th Symposium on Operating Systems Principles, pp. 33-48, October 1983).
Tuesday May 16
- No readings. Finish your assignment and in class we will be benchmarking your web servers and discussing your designs and their performance.
Tuesday May 23 (summaries required for all)
- Why Threads Are A Bad Idea (for most purposes)
- Flash: An Efficient and Portable Web Server.
- Event-driven programming for robust software,
- Why Events Are A Bad Idea (for High-concurrency Servers)
Tuesday May 30 (summaries required for all)
- SEDA: An Architecture for Well-Conditioned, Scalable Internet Services
- Multiprocessor Support for Event-Driven Programs
- Capriccio: Scalable Threads for Internet Services
Tuesday June 6 (summaries required for all)
- Measuring the Capacity of a Web Server
- httperf---A Tool for Measuring Web Server Performance
- Open Versus Closed: A Cautionary Tale
Tuesday June 13 (summaries required for all)
- IO-Lite: A Unified I/O Buffering and Caching System
- Kernel Support for Faster Web Proxies
- Lazy Asynchronous I/O For Event-Driven Servers
Tuesday June 20 (summaries required for all)
- Scheduler activations: effective kernel support for the user-level management of parallelism
- Cost-Aware WWW Proxy Caching Algorithms
- Web Server Workload Characterization: The Search for Invariants
Tuesday June 27 (summaries required for all)
- Web Workload Characterization: Ten Years Later
- Tracking the Evolution of Web Traffic: 1995-2003
- Flux: A Language for Programming High-Performance Servers
Tuesday July 4 (summaries required for all but the Article in Macleans)
- But what do you want a Ph.D. for? by Andrew Potter, Macleans Magazine, June 26, 2006
- Evaluating the Impact of Simultaneous Multithreading on Network Servers Using Real Hardware
- An Evaluation of Network Stack Parallelization Strategies in Modern Operating Systems
- Evaluating Network Processing Efficiency with Processor Partitioning and Asynchronous I/O
Tuesday July 11 (summaries required for all but "Scalable Web Server Clustering Technologies")
- Cluster-Based Scalable Network Services
- Lessons from Giant-Scale Services
- Web Search for a Planet: The Google Cluster Architecture
- Scalable Web Server Clustering Technologies,
Tuesday July 18 (summaries required for all)
- Understanding and Addressing Blocking-Induced Network Server Latency
- Connection Conditioning: Architecture-Independent Support for Simple, Robust Servers

Last modified: Wed Jun 21 22:08:43 EDT 2006