Jet-Lag Java Heterogeneous Log Analysis on the Grid: Architecture, Implementation and Performance Evaluation

This paper describes a new Grid application which performs analysis of log files conforming to several different standard formats. In order to exploit the potential execution parallelism of the Grid, we designed and developed a software architecture which splits log files into smaller fragments which are then analyzed independently, possibly in parallel, using computing resources available in a Grid environment. We present the distributed architecture of our Grid log analyzer and describe Jet-Lag, our first Java implementation. We then propose an experimental setting to test Jet-Lag functionality and performance on a DataGrid test-bed. First experimental results show that Grid log analysis with the proposed software architecture is feasible and effective. Performance, as expected for a first Java prototype implementation, is not yet optimal, but main bottlenecks have been identified. A considerable amount of experimental data has been collected that will provide reference material for further investigation.