PACMan: Performance Aware Virtual Machine Consolidation

  • Alan Roytman
  • Aman Kansal
  • Sriram Govindan
  • Jie Liu

10th International Conference on Autonomic Computing (ICAC) |

Published by USENIX

Consolidation of multiple workloads, encapsulated in virtual machines (VMs), can significantly improve efficiency in cloud infrastructures. But consolidation also introduces contention in shared resources such as the memory hierarchy, leading to degraded VM performance. To avoid such degradation, the current practice is to not pack VMs tightly and leave a large fraction of server resource unused. This is wasteful. We present a system that consolidates VMs such that performance degradation is within a tunable bound while minimizing unused resources. The problem of selecting the most suitable VM combinations is NP-Complete and our system employs a practical method that performs provably close to the optimal. In some scenarios resource efficiency may trump performance and for this case our system implements a technique that maximizes performance while not leaving any resource unused. Experimental results show that the proposed system realizes over 30% savings in energy costs and up to 52% reduction in performance degradation compared to consolidation algorithms that do not consider degradation.