1 / 18

Artemis

Artemis. Distributed system. Hunting for Bugs with Artemis. Dryad Overview. Logs. System Architecture. Data Collection. Data collection. Database. View. GUI. Plug-ins. GUI. Plug-ins. Conclusions. Hunting for Bugs with Artemis. Gabriela F. Creţu-Ciocârlie Mihai Budiu

shadi
Télécharger la présentation

Artemis

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. Artemis Distributed system Hunting for Bugs with Artemis Dryad Overview Logs System Architecture Data Collection Data collection Database View GUI Plug-ins GUI Plug-ins Conclusions

  2. Hunting for Bugswith Artemis Gabriela F. Creţu-Ciocârlie Mihai Budiu Moises Goldszmidt Microsoft Research, Silicon Valley WASL 2008 This presentation is built and should be viewed with pptPlex: http://www.officelabs.com/projects/pptPlex/Pages

  3. Artemis Goal One-stop shop for performance analysis of distributed systems

  4. Principles • 1) Modular: Separate generic from application specific parts • 2) Extensible: add new analyses via plug-ins • 3) Interactive: human expert part of the analysis loop

  5. Distributed system Distributed Logs Data collection Database Local View GUI Plug-ins

  6. Distributed system Application-Specific Logs Data collection Generic Database View GUI Plug-ins

  7. Dryad Application Structure Inputfiles Channels Stage Outputfiles sort grep awk sed perl sort grep awk sed grep sort Vertices

  8. Dryad System Architecture data plane job schedule V V V Serv Serv Serv control plane Job manager cluster

  9. Text Binary XML Perfmon Data Text Binary XML Perfmon Text Binary XML Perfmon 10GB-1TB Copy DryadLINQ application Persisted data Parse Filter Aggregate 100MB-1GB

  10. Machine Utilization Plug-in

  11. Complex statistics: HiLighter plug-in Key Performance Indicator Binary search overlogistic regression with L1 regularization Correlated metrics Metrics

  12. Interactive Analysis KPI Selection Feature Computation Visualization Hilighter

  13. Conclusions Automatic diagnosis Goal Statistical analyses Feature extraction Artemistoday Summarization Raw data Distributed system

More Related