Bonn 2020 – wissenschaftliches Programm
Die DPG-Frühjahrstagung in Bonn musste abgesagt werden! Lesen Sie mehr ...
T 25.1: Vortrag
Dienstag, 31. März 2020, 17:00–17:15, H-HS V
Simulating distributed caching for end user analyses — René Caspart, R. Florian von Cube, •Tabea Feßenbecker, Max Fischer, Manuel Giffels, Christoph Heidecker, Maximilian Horzela, Eileen Kühn, Günter Quast, and Matthias Schnepf — Karlsruhe Institute of Technology
With the start of the High-Luminosity LHC era in the near future, the massive amount of data processed in HEP user analyses results in challenges for network and storage infrastructures. In order to face these challenges developing concepts to use existing computing infrastructures more efficiently becomes essential.
Bringing data close to the computing resources is a promising approach to overcome throughput limitations and improve the overall performance. One way to achieve this is the placement of caches close to the computing resource. In order to efficiently use caches in a heterogeneous, distributed computing infrastructure an adapted operation concept needs to be developed.
Our goal at KIT is to develop a concept for facilities focused on end-user analysis. Therefore, we monitor our system and create a simulation to study the effects of different caching concepts. This allows to investigate the possibilities that distributed caching offers to computing facilities designed for end-user analysis.
In this contribution, we give an overview of the simulation tool and discuss different cache operation concepts of distributed caching.