An Energy-Efficient VM Prediction and Migration Framework for Overcommitted Clouds
Abstract
The An Energy-Efficient VM Prediction and Migration Framework for Overcommitted Clouds makes great energy savings by minimizing physical machine (PM) overload occurrences through monitoring and prediction of the use of VM resources, and reducing the number of active PMs through efficient VM migration and placement. Using real Google data consisting of 29-day traces collected from a cluster containing more than 12 K PMs, we show that our proposed framework outperforms existing overload avoidance techniques and prior VM migration strategies by reducing the number of unforeseen overloads, minimizing overload migration, increasing resource utilization and reducing cloud energy consumption.
System Configuration
Conclusion
An energy-efficient VM Prediction and Migration Framework for Overcommitted Clouds, We propose an integrated energy-efficient, predictive-based VM placement and migration framework for cloud resource allocation with overcommitment. We show that our proposed framework reduces performance degradation caused by overloads while also reducing the number of PMs needed to be ON and overhead migration, resulting in significant energy savings. All of our findings are supported by evaluations and comparative studies with existing techniques conducted from a Google cluster on real traces.
