Research Area:  Cloud Computing

Abstract:

Since service level agreement (SLA) is essentially used to maintain reliable quality of service between cloud providers and clients in cloud environment, there has been a growing effort in reducing power consumption while complying with the SLA by maximizing physical machine (PM)-level utilization and load balancing techniques in infrastructure as a service. However, with the recent introduction of container as a service by cloud providers, containers are increasingly popular and will become the major deployment model in the cloud environment and specifically in platform as a service. Therefore, reducing power consumption while complying with the SLA at virtual machine (VM)-level becomes essential. In this context, we exploit a container consolidation scheme with usage prediction to achieve the above objectives. To obtain a reliable characterization of overutilized and underutilized PMs, our scheme jointly exploits the current and predicted CPU utilization based on local history of the considered PMs in the process of the container consolidation. We demonstrate our solution through simulations on real workloads. The experimental results show that the container consolidation scheme with usage prediction reduces the power consumption, number of container migrations, and average number of active VMs while complying with the SLA.

Keywords:  

Author(s) Name:  Jialei Liu, Shangguang Wang, Ao Zhou, Jinliang Xu & Fangchun Yang

Publisher name:  Springer

DOI:  10.1007/s11704-018-7172-3

Volume Information:  volume 14, pages 42–52 (2020)

Paper Link:   https://link.springer.com/article/10.1007%2Fs11704-018-7172-3