Münster 2017 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

T: Fachverband Teilchenphysik

T 66: Grid-Computing

T 66.6: Talk

Tuesday, March 28, 2017, 18:05–18:20, JUR 372

Advantages of coordinated caching for run 2 analyses — •Christoph Heidecker, Max Fischer, Manuel Giffels, and Günter Quast — Karlsruher Institut für Technologie

The current run 2 of the LHC delivers an enormous amount of data waiting for further processing. This includes not only big challenges to the storing capabilities, but also to the distribution of data to the end-user for further analyses. At the KIT, a coordinated caching on SSDs was designed to process large datasets on short turn around cycles. The typical workflow of an analysis at high energy physics requires an iterated execution of the analysis tools. Usually, this is done via a batch-system which allows a parallel execution on different worker nodes. The High Throughput Data Acquisition system caches the necessary datasets directly after the first execution on SSDs located at the worker nodes. This allows an faster processing of data from the second execution of the analysis. Hence, the processing time is reduced dramatically which allows a prompt deployment of new results. The iterated deployment of the jet energy corrections allowed an detailed testing of the caching system with an increasing amount of data. In this context, the developments and experiences of the Karlsruhe workgroup using this coordinated caching will be presented.