Companies move their applications and services to the cloud. A se vice level agreement (SLA) ensures the user high uptime and small response times from the data center provider. To be cost efficient, the service provider has to make trade-off calculations between SLA violation penalties and additional hardware. Especially peak times and co-scheduled applications influence the performance. By creating interferences with shared resources, peak times can be simulated. Each resource can be separately tested for potential SLA violations. An SLA violation means, that the system needs a higher resource limit, no violation means that the provider can reduce the resource.

The Resource Contention Tool is a workload suite with small workloads, that occupy one shared resource for a given amount of time and intensity. By occupying one shared resource, we create a resource contention with the service or application on the data center. The tool consists of carefully crafted workloads for CPU, memory, cache and storage, and can easily be extended with additional workloads. Changing the intensity of workloads allows to observe the sensitivity of applications towards resource interferences.

Titel Resource Contention of Competing Processes in Parallel Systems
Verfasst von Christian Würz
Serien-Detail Masterarbeit
Verlag Hasso-Plattner-Institut an der Universität Potsdam
Datum 20. Oktober 2017
Seitenzahl 61
HinzugefĂĽgt am 13. April 2021
HinzugefĂĽgt von max.plauth
Buch-Verleih verfĂĽgbar
PDF-Download Fachgebiets-Angehörige+ max.plauth