Evaluating DataSeries for use with a Job Centric Monitoring Service

Abstract

The role of monitoring data in IT environments is changing. Once reserved primarily for debugging purposes, monitoring data is increasingly being integrated into the daily use and management of IT systems. These new uses place different requirements on the storage and access to the monitoring data. We believe that such requirements are not met by existing data providers. In this paper we propose using DataSeries, a generic trace format developed at HP Labs, to address these emerging requirements. We examine the data access and storage requirements of a Job Centric Monitoring (JCM) service developed at the Grid Research Centre (GRC), and explain how DataSeries could fulfill these requirements. We then explore the performance of DataSeries for both the access and storage requirements, using empirical data to help us illustrate the tradeoffs that exist. As future work the GRC intends to integrate DataSeries into the JCM service.