Leveraging Resource Utilization Patterns in Cluster Management

EDIC Candidacy Exam
Exam president: Prof. Katerina Argyraki
Thesis advisor: Prof. Willy Zwaenepoel
Co-examiner: Prof. Rachid Guerraoui

Abstract
Efficient server resource utilization remains a challenge for large-scale cloud operators. Workload co-location and server resource over-commitment are some approaches to increase resource utilization and reduce costs in datacenters. Special care must be taken to not impact application performance.
We review three cluster management systems, namely Borg, Quasar, and History-based Harvesting. These systems rely on resource utilization patterns to perform workload co-location and over-commitment without reducing performance. Finally, we present future research directions to quantify and exploit resource usage predictability in cluster management decisions.​

Background papers
Large-scale cluster management at Google with Borg by Verma, A., et al.

Quasar: Resource-Efficient and QoS-Aware Cluster Management, by  Delimitrou, C., and Kozyrakis, C.

History-Based Harvesting of Spare Cycles and Storage in Large-Scale Datacenters Zhang, Z.,  Prekas, G., et al.