Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 99: Grid-Computing II

T 99.5: Vortrag

Donnerstag, 12. März 2015, 17:45–18:00, K.11.10 (K8)

The Smart Grid Monitoring System — Gen Kawamura, •Erekle Magradze, Haykuhi Musheghyan, Jordi Nadal, Arnulf Quadt, and Gerhard Rzehorz — II. Physikalisches Institut, Georg-August-Universität Göttingen, Göttingen, Deutschland

Large digital service providers like grid or cloud resource centers have the distributed structure, which consists of a number of computing and storage centers - the sites. Each of these centers have the Service Oriented Architecture (SOA) in order to simplify and standardize the communication and service provisioning to the end users and to other services. High availability, reliability and serviceability of such distributed infrastructures depend on efficient management of each site. In order to avoid the approaching issues affecting the service availability, a proactive management is essential. The key aspect in proactive management is the prediction of any service failure, which is impossible without definition of a proper metric for it. In this study, we present an application of Service Response Time (SRT) as a metric for the service failure and techniques allowing to predict it up to eight hours in advance. An approach for forecasting is based on Adaptive Network based Fuzzy Inference System (ANFIS), which shows more than 90% efficiency according to the ten fold cross validation results for two case studies. Implementation of such techniques for large computing facility management is important to increase the computing infrastructure availability, reliability and serviceability.