ITR: Resource Predictors from Application Signatures in High Energy Physics

Harvard University, Cambridge MA

Investigators

Abstract

This proposal requests support for investigators from Harvard and Northwestern Universities for an investigation of the creation and use of application signatures from the ATLAS high energy physics experiment, at the Large Hadron Collider (LHC) at CERN, Switzerland, as estimators for resource usage and adaptation for computing on a Grid. The experiment will be working with the largest data sets ever used in a physics experiment and will involve physicists distributed world wide. The access to these data and the analyses of them will require a grid network. Rather than just transfer the large data sets around the world, the scientists need to be able to regenerate the data by taking a limited signature of a specific data set, called an application signature, and invoke methods to cause a physical materialization. The group proposes to develop application signatures for jobs in the ATLAS experiment; to develop tools to estimate the execution times required to complete computing tasks based on a limited set of application signatures; to use this information to help schedule tasks on a set of grid resources; to find the most efficient way to solve a particular ATLAS analysis problem; and to evaluate their results in terms of generalization to other applications.