Web object life measurement using Squid log file

Abstract

Web object life was directly and costly derived by directly polling at original Web servers. A new Web object life measurement is proposed using a proxy and cache server log file. Each object-request record is classified into life-related types, Life_Hit, Life_Miss, Life_Unknown or Life_Ignore. An object life is calculated using Life_Hit and Life_Miss's timestamps. This indirect measurement has been analyzed for its error and accuracy. Error and accuracy are better if there are more life-related requests for an object. Squid's access log file is used, and its results codes are mapped to life-related types. All requested object lives are calculated and shown in distribution graphs with average accuracy indicated.