Göttingen 2012 – scientific programme

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

T: Fachverband Teilchenphysik

T 76: Computing 1

T 76.7: Talk

Wednesday, February 29, 2012, 18:15–18:30, VG 0.111

Learning and Decision-making Monitoring System for Grid Sites — Georg Jahn, •Erekle Magradze, Jörg Meyer, Arnulf Quadt, and Christian Wehrberger — II. Physikalisches Institut Georg-August Universität Göttingen, Deutschland

Large scale computing systems, such as WLCG have a very complex infrastructure with more than hundred scientific computing centers - called sites. The sites define the computational and storage power of WLCG. Each computing center by itself is a complex system with dif- ferent components and each of them is monitored for characteristics and status of resources. Monitoring of site systems and central services of WLCG is a lively research area.

Nowadays there are numbers of monitoring tools dedicated for retriev- ing the statuses of site infrastructure and central services of WLCG. As a result, there is a big flow of monitoring information from the tools, which is analysed daily by site and central services administra- tors. Automatisation of analytical processing of the information from various monitoring sources and consequently issuing appropriate ac- tions to solve particular problems appearing at sites, or among the central services, is a promising way to increase the site performance and availability.

A first meta-monitoring system with features of learning and decision- making for the administration of sites will be presented. The tool is being developed and tested at the WLCG ATLAS Tier-2 center Goe- Grid. The technical approach based on Artificial Neural Networks and Fuzzy Sets Theory will be discussed.