Electrical Engineering
      and Computer Sciences

Electrical Engineering and Computer Sciences


UC Berkeley


2009 Research Summary

Virtual Cluster Management

View Current Project Information

Gunho Lee, Byung-Gon Chun1 and Randy H. Katz

When it comes to cloud computing, virtual machines play an important role. By using virtual machines as the unit of management, we can deploy and scale them up quickly and easily. As a matter of fact, most of cloud computing (or utility computing) providers adapt virtual machines to deploy the user's application on their datacenters.

In many cases, resources guaranteed (or assigned) for a virtual machine are much less than that of a physical machine. Therefore, multiple virtual machines would be placed on single physical machine to maximize resource utilization and cost performance. However, the way we place virtual machines would significantly affect the overall performance. According to the characteristics of workload runs on those virtual machines, we would classify the relationship between them as interference, orthogonality, and affirnity.

In this project, we will investigate the relationship between various workloads and come up with an efficient strategy to map virtual machines to physical machines.

1Intel Research Berkeley