EmbracingFailureAvailabilityviaRepair-centricDesign.pptVIP

Embracing Failure:Availability via Repair-centric Design;Contributions;Outline;Motivation for a new philosophy;Traditional HA vs. Internet reality;Facts of life;Repair-centric hypothesis;Repair-centric systems;Context: repair-centric design;Outline;Research Plan;Context: application;Context: implementation platform ;Staged research plan;Stage 1: Undo;Undo examples;Undo context;Undo implementation;Undo implementation (2);Undo issues;Stage 2: Online verification;Issues in online verification;Stage 3: Exercising repair;Exercising repair: approach;Issues in exercising repair;Stage 4: Diagnosis aids;Diagnosis aids;Outline;Evaluation plan;Availability benchmarks quantify system behavior under failures and maintenance They require a realistic workload for the system quality of service metrics and tools to measure them fault-injection to simulate failures human operators to perform repairs;Availability benchmarks for email;Fault injection;Evaluation: human aspects;Summary;Timeline;End;Human error rate experiments;What causes un-availability?;Traditional HA vs. repair-centric;Assumptions;Details: application spectrum;Context: undo;Context: exercising repair;Context: online verification;Context: diagnosis;What we’re NOT trying to do;Ninja details;Context: repair-centric design