Optimized cloud placement of virtual clusters using biased importance sampling

We introduce an algorithm for the placement of constrained, networked virtual clusters in the cloud, that is based on importance sampling (also known as cross-entropy). Rather than using a straightforward implementation of such a technique, which proved inefficient, we considerably enhance the method by biasing the sampling process to incorporate communication needs and other constraints of placement requests to yield an efficient algorithm that is linear in the size of the cloud. We investigate the quality of the results of using our algorithm on a simulated cloud.