Interference Aware Container Orchestration in Kubernetes Cluster

Abstract

Nowadays, there is an ever-increasing number of workloads pushed and executed on the Cloud. To effectively serve and manage these huge computational demands, data center operators and cloud providers have embraced workload co-location and multi-tenancy as first class system design concern. In addition, the continuous advancements in the computers’ hardware technology have led to a heterogeneous pool of systems lying under data center environments. Current state-of-the-art schedulers and orchestrators rely on typical metrics, such as CPU or memory utilization, for placing incoming workloads on the available pool of resources, thus, not taking into consideration the interference effects each task cause, when co-located with others, as well as the impact of systems’ underlying diversity on the performance. In this thesis, we design an interference- and heterogeneity- aware cloud orchestrator, able to efficiently schedule applications arriving at a data center on a pool of available resources. We showcase the impact of applying stress on different shared resources of two heterogeneous server systems and we propose an indicator that depicts the state of the system based on these observations. We integrate our solution with Kubernetes, one of the most widely used cloud orchestration frameworks nowadays, and we show that we can achieve higher performance compared to its default scheduler, for a variety of cloud representative workloads.