An Energy and Performance Aware Consolidation Technique for Containerized Datacenters

Cloud datacenters have become a backbone for today’s business and economy, which are the fastest-growing electricity consumers, globally. Numerous studies suggest that <inline-formula><tex-math notation="LaTeX">$\sim$</tex-math></inline-formula> 30% of the US datacenters are comatose and the others are grossly less-utilized, which make it possible to save energy through resource consolidation techniques. However, consolidation comprises migrations that are expensive in terms of energy consumption and performance degradation, which is mostly not accounted for in many existing models, and, possibly, it could be more energy and performance efficient not to consolidate. In this paper, we investigate how migration decisions should be taken so that the migration cost is recovered, as only when migration cost has been recovered and performance is guaranteed, will energy start to be saved. We demonstrate through several experiments, using the Google workload data for 12,583 hosts and approximately one million tasks that belong to three different kinds of workload, how different allocation policies, combined with various migration approaches, will impact on datacenter’s energy and performance efficiencies. Using several plausible assumptions for containerised datacenter set-up, we suggest, that a combination of the proposed energy-performance-aware allocation ( <small>Epc-Fu</small> ) and migration ( <small>Cper</small> ) techniques, and migrating relatively long-running containers only, offers for ideal energy and performance efficiencies.