Cost-effective provisioning and scheduling of deadline-constrained applications in hybrid clouds

In order to meet distributed application deadlines, Resource Management Systems (RMSs) have to utilize additional resources from public Cloud providers when in-house resources cannot cope with the demand of the applications. As a means to enable this feature, called Cloud Bursting, the RMS has to be able to determine when, how many, and for how long such resources are required and provision them dynamically. The RMS has also to determine which tasks will be executed on them and in which order they will be submitted (scheduling). Current approaches for dynamic provisioning of Cloud resources operate at a per-job level, ignoring characteristics of the whole organization workload, which leads to inefficient utilization of Cloud resources. This paper presents an architecture for coordinated dynamic provisioning and scheduling that is able to cost-effectively complete applications within their deadlines by considering the whole organization workload at individual tasks level when making decisions and an accounting mechanism to determine the share of the cost of utilization of public Cloud resources to be assigned to each user. Experimental results show that the proposed strategy can reduce the total utilization of public Cloud services by up to 20% without any impact in the capacity of meeting application deadlines.