Optimizing resource utilization in private clouds with an opportunistic approach

From the perspective of a typical cloud user, clouds are a source of unlimited computational resources that can be contracted as needed. As a consequence, users are relieved from the task of planning the capacity of their infrastructure. In practice, the problem of capacity planning has simply shifted from the cloud users to the cloud providers. The main challenge of cloud providers is, therefore, to reduce as much as possible the idleness of their infrastructure, at the same time that they appropriately support their variable workload. One approach used by public cloud providers to reduce the infrastructure idleness is to offer resources with a degraded quality of service. However, in private clouds, the idea of exploiting idle capacity has been little explored. In this demo, we present a system that opportunistically explores idle capacity in private cloud providers. The best effort service provided by resources exploited opportunistically can be efficiently used by users that need to run bag-of-task jobs as fast as possible-a workload that is becoming more and more common nowadays.