Automatic Failure-Path Inference: A Generic Introspection Technique for Internet Applications

  • Authors:

  • George Candea;Mauricio Delgado;Michael Chen;Armando Fox

  • Affiliations:

  • -;-;-;-

  • Venue:
  • WIAPP '03 Proceedings of the The Third IEEE Workshop on Internet Applications
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

Automatic Failure-Path Inference (AFPI) is anapplication-generic, automatic technique for dynamicallydiscovering the failure dependency graphs of componentizedInternet applications. AFPI's first phase is invasive,and relies on controlled fault injection to determine failurepropagation; this phase requires no a priori knowledgeof the application and takes on the order of hours torun. Once the system is deployed in production, the second,non-invasive phase of AFPI passively monitors thesystem, and updates the dependency graph as new failuresare observed. This process is a good match for theperpetually-evolving software found in Internet systems;since no performance overhead is introduced, AFPI isfeasible for live systems. We applied AFPI to J2EE andtested it by injecting Java exceptions into an e-commerceapplication and an online auction service. The resultinggraphs of exception propagation are more detailed andaccurate than what could be derived by time-consumingmanual inspection or analysis of readily-available staticapplication descriptions.