Prediction of Information System Availability in Mission Critical and Business Critical Applications

摘要

One of the most important attributes ofrnon-line computer systems that are performing missionrnor business critical applications is availability. SystemrnEngineers are often called upon to predict the reliabilityrnof such systems as part of proposal preparation,rnarchitecture definition, design reviews, and operation.rnHowever, traditional modeling techniques are incapablernof handling integrated hardware and software systemsrnwith multiple states and redundancy. This paperrndescribes how Markov Modeling and Reliability BlockrnDiagrams can be used together to model a large on-linerninformation system and develop the answers to strategicrnquestions on the configuration and operation of highrnavailability computing systems. The analyses arernperformed using MEADEP, a reliability analysis toolrncapable of hierarchical modeling and integratingrnMarkov and block diagram techniques. In the examplern3-tier architecture e-commerce site described in thisrnpaper, it is shown that (a) the most frequently failingrnsubsystem is not necessarily the availability bottleneck,rnand (b) that restoration time is often a more importantrnparameter than availability when attempting tornmaximize system throughput.