1 / 29

Operating System Support for Application-Specific Speculation

Operating System Support for Application-Specific Speculation. Benjamin Wester Peter Chen and Jason Flinn University of Michigan. Speculative Execution. Sequential dependent tasks Predict results of Task A to break dependence Execute Task B in parallel Isolate all effects

kaiser
Download Presentation

Operating System Support for Application-Specific Speculation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Operating System Support for Application-Specific Speculation Benjamin Wester Peter Chen and Jason Flinn University of Michigan

  2. Speculative Execution • Sequential dependent tasks • Predict results of Task Ato break dependence • Execute Task B in parallel • Isolate all effects • Correct prediction: commit • Wrong prediction: abort TIME Predict A A B B EuroSys 2011

  3. Speculation Everywhere! • Discrete event simulation • I/O prefetching • Distributed shared memory • Distributed file systems • Deadlock detection • Remote displays • Web page pre-rendering EuroSys 2011

  4. Speculation as a Service to Apps How is this system designed? In what ways can it be customized for an app? How can those customizations be specified? EuroSys 2011

  5. Outline • Introduction • Designing Speculation as a Service • Implementation • Evaluation • Conclusion EuroSys 2011

  6. Design 1: In-App Speculation • Complete semantic info • Predict arbitrary app operations • Safe operations allowed • No reuse: significant development needed • Scope is limited: unsafe operations block App 2 App 1 Data Data EuroSys 2011

  7. Design 2: Generic OS Speculation • Apps need no modifications • Wide scope: unsafe operations taint • Lacks semantic understanding of app • Predict system calls only • Handle application conservatively OS App 1 App 2 EuroSys 2011

  8. Separate Mechanism and Policy Mechanismimplements isolation Policydescribes customizations Best of both extremes • Mechanism built in OS • Common implementation • Wide scope • Policy specified in Applications • Expose semantic information EuroSys 2011

  9. Design 3: Expose Predictions • Predict arbitrary app operations • Reuse OS mechanism(with app assistance) • Wide scope for taint propagation • Limited semantic info • Speculative external output never allowed • Commit on identicalresults OS App 2 App 1 EuroSys 2011

  10. Design 4: Expose Safety • Predict arbitrary app operations • Reuse OS mechanism(with app assistance) • Wide scope for taint propagation • More semantic info • Allow safeoutput • Commit on equivalentresults OS App 2 App 1 EuroSys 2011

  11. Customizable Policy • Creation • What tasks are predictable • How to predict them • Output • What output is safe to allow • Commit • Which results are acceptable to commit EuroSys 2011

  12. Outline • Introduction • Designing Speculation as an OS Service • Implementation • Evaluation • Conclusion EuroSys 2011

  13. Implementation • Mechanism built in OS • Based on Speculator kernel • Checkpoints & logs processes, files, IPC, etc. • Policies expressed using system call API EuroSys 2011

  14. spec_fork() TIME spec_fork predict A A B Control Process Speculative Process commit abort C B EuroSys 2011

  15. API Example int main() { int x; int prediction = get_prediction(); if (spec_fork() == SPECULATIVE) { x = prediction; } else { x = slow_function(); if (equiv(x, prediction)) commit(); else abort(); } set_output_policy(stdout, ALLOW); printf(“%d”, x); } Creation Policy Commit Policy Output Policy EuroSys 2011

  16. Outline • Introduction • Designing Speculation as an OS Service • Implementation • Evaluation • Conclusion EuroSys 2011

  17. Evaluation Can apps effectively use APIto increase parallelism? Case studies • Predictive application launching in Bash • SSL certificate checks in Firefox • Replicated service in PBFT-CS EuroSys 2011

  18. App 1: Predictive Launching in Bash …finished bwester $ ▌ Run program Creation Policy: Use machine learning to predict user input TIME …finished bwester $ grep foo –r . ▌ Check command grep foo –r . Commit Policy: Normalize user input before comparison Output Policy: Safe X11 Messages: Allow EuroSys 2011

  19. How Much Work Can Be Hidden? Non-Speculative Speculative 45 Prompt Prompt TIME Spec. Executecmd Get cmd Get cmd Execute Executecmd EuroSys 2011

  20. How Much Work Can Be Hidden? Up to 86% hidden Non-Speculative Speculative 45 Prompt Prompt TIME Spec. Executecmd Get cmd Get cmd Execute Executecmd EuroSys 2011

  21. App 2: Firefox SSL Connections Creation Policy: Predict certificate is valid Web Server Open https:// TIME Output Policy: Alow SSL connection Check cert. Validation Server Validate Get session key Request page Output Policy: Block private data GET /index.html?id=0123 Done EuroSys 2011

  22. Connection Latency Hidden? Non-Speculative Speculative https:// https:// TIME Check cert. Check cert. Session key Session key Validate Request Request Done Done EuroSys 2011

  23. Connection Latency Hidden? Non-Speculative Speculative Avg 60ms hidden https:// https:// TIME Check cert. Check cert. Session key Session key Validate Request Request Done Done EuroSys 2011

  24. App 3: PBFT-CS Protocol Creation Policy: Agree on 1st reply Request Distributed Replicated Service TIME Send Request (Reply, sig1) Check reply (Reply, sig2) Check reply Work… Output Policy: PBFT-CS Messages: Allow EuroSys 2011

  25. Improved Client Throughput? Non-Speculative Speculative TIME Request1 Request1 1st Reply 1st Reply Request2 Null requests 1st Reply Verify Verify Request2 1st Reply EuroSys 2011

  26. Improved Client Throughput? 1.9x Throughput Non-Speculative Speculative TIME Request1 Request1 1st Reply 1st Reply Request2 Null requests 1st Reply Verify Verify Request2 1st Reply EuroSys 2011

  27. Cost of Generic Mechanism App-specific Mech Shared OS Mech Non-speculative Null requests EuroSys 2011

  28. Cost of Generic Mechanism App-specific:8% faster App-specific Mech Shared OS Mech Non-speculative Null requests EuroSys 2011

  29. Conclusion • Mechanism • Common: checkpoints, output buffering, taint propagation • Implemented in OS • Policy • App-specific: Controls creation, output, and commit • Implemented in applications • Demonstrated with 3 case studies • Improved parallelism • Small overhead relative to app-specific mechanism Questions? EuroSys 2011

More Related