CS 290I -- Practical Performance Evaluation (Winter 98)

(aka Experiment Design and Measurement)
Anurag Acharya, acha@cs.ucsb.edu
Mon/Wed 11:00--12:50, 1401 Phelps
Office hours: by email appointment
Enrollment code: 17285

Short description

The purpose of this course is to outfit a practical systems designer with the tools of the trade. I plan to cover practical concerns such as basic experiment design, workload selection, measurement tools, simulation techniques, and perturbation effects as well as basic statistics and probability theory that underlie all practical evaluation.

A large part of this course will be an evaluation project. Projects will be individual. You can propose your own projects, possibly something from your research domain or another class you are taking. In addition, I have a set of potential projects in several areas (compilers, operating systems, networking, web servers, architecture, memory management etc). Other important parts are critiques of evaluation in published papers and class participation.

Topics covered include: system metrics, workload selection, monitoring techniques, benchmarking, data manipulation and presentation, testing and comparing data, statistical metrics, factor experiments, designing and controlling simulations and distribution generation.

Long description

Project reports

  • Karel Driesen, Improving Indirect Branch Prediction With Source and Arity Based Classification and Cascaded Prediction
  • Hakan Ferhatosmanoglu, Performance Evaluation of Declustering Techniques for Similarity Searches
  • Huican Zhu and Kai Shen, Impact of Cache on Trie Search
  • Andrew Rutz, Performance Evaluation of Two Offline Java Compilers
  • Ben Smith, Cooperative Caching of Dynamic Content on a Distributed Web Server
  • Suk Lee, Performance Evaluation of Different Log-based Rollback-recovery Protocols
  • David Watson, Memory Performance of Conservative Garbage Collection for Long-running Programs
  • Mandar Raje and Sean Brydon, Performance Evaluation of Web Servers.
  • Maximilian Ibel, Soo Kim, Michael Schmitt, Adaptive Communication Strategies for NOWs.