Resource Signal Prediction and Its Application to Real-time Scheduling Advisors

Peter A. Dinda, School of Computer Science, Carnegie Mellon University

Successfully defended on 2/14/2000

Dissertation

  • Dissertation: P. Dinda, Resource Signal Prediction and Its Application to Real-time Scheduling Advisors, Technical Report CMU-CS-00-131, School of Computer Science, Carnegie Mellon University, February, 2000. abstract postscript pdf

  • Summary: P. Dinda, Resource Signal Prediction and Its Application to Real-time Scheduling Advisors, February, 2000. postscript

  • Software

  • RPS Toolkit: Tools for building on-line and off-line resource prediction systems. Described in CMU-CS-99-138, available below.

  • Playload: Tools for reconstructing workloads from host load traces. Described in LCR00 paper, soon to be available below.

  • Load Trace Archive: A large collection of fine grain host load traces.

  • Raw Testcase Database : A large number of randomized testcases that were used to evaluate different aspects of the thesis system. These are currently only available by special request.

  • Papers

  • P. Dinda, D. O'Hallaron, Realistic CPU Workloads Through Host Load Trace Playback, In Proc. 5th Workshop on Languages, Compilers, and Run-time Systems for Scalable Computers, Rochester, NT, May, 2000, TO APPEAR. abstract postscript

  • P. Dinda, D. O'Hallaron, Host Load Prediction Using Linear Models , Cluster Computing, To Appear (This is the journal version of the HPDC '99 paper. It describes the study in more detail and gives performace data for the on-line host load prediction system.)

  • P. Dinda, D. O'Hallaron, An Extensible Toolkit for Resource Prediction In Distributed Systems, technical report CMU-CS-99-138, School of Computer Science, Carnegie Mellon University, July, 1999. abstract, postscript

  • P. Dinda, D. O'Hallaron, An Evaluation of Linear Models for Host Load Prediction, Proceedings of the 8th IEEE Symposium on High-Performance Distributed Computing (HPDC-8) Aug 1999, Redondo Beach, CA. abstract (with BibTex entry), postscript, pdf.

  • P. Dinda, B. Lowekamp, L. Kallivokas, D. O'Hallaron, The Case for Prediction-based Best-effort Real-time Systems, In Proc. of the 7th International Workshop on Parallel and Distributed Real-Time Systems (WPDRTS 1999), San Juan, April, 1999, pp. 309--318 abstract postscript (CMU-CS-98-174 is an extended version)

  • P. Dinda, The Statistical Properties of Host Load , Scientific Programming, 7:3-4, Fall, 1999. (This is a vastly extended version of the study published in LCR 98 (below) and presents results for a second set of machines. This is currently as CMU-CS-98-175 abstract, postscript )

  • Other papers by Peter A. Dinda

  • Talks

  • Defense Talk - will be available shortly

  • The Case for Prediction-based Best-effort Real-time - WPDRTS '99 4/12/99 powerpoint

  • A Prediction Service for Remos and QuO - Systems Seminar 3/1/99 powerpoint

  • Load Trace Replay - Systems Seminar 11/23/98 powerpoint

  • Load Prediction for Best-effort Real Time - ARPA Quorum San Diego PI Meeting - August, 1999 powerpoint

  • Statistical Properties of Host Load - LCR '98 5/30/98 powerpoint

  • Responsive Interactive Applications by Dynamic Mapping of Activation Trees - BBN Invited Talk 2/20/98 powerpoint

  • Markov Modeling of Host Load - Systems Seminar 2/9/98 powerpoint

  • Proposal

  • Abstract
  • Document
  • Talk (presented 1/29/98)
  • Other Related Items

  • Load trace spectrogram ( axp7.psc.edu (8/19/97, 24 hours)
  • Load trace self-similarity (All estimators), (Mean +/- sdev)

  • Peter A Dinda
    Last modified: Tue Aug 29 14:47:08 EDT 2000