Scheduling of data-intensive workloads in a brokered virtualized environment

Abstract

Providing performance predictability guarantees is increasingly important in cloud platforms, especially for data-intensive applications, for which performance depends greatly on the available rates of data transfer between the various computing/storage hosts underlying the virtualized resources assigned to the application. With the increased prevalence of brokerage services in cloud platforms, there is a need for resource management solutions that consider the brokered nature of these workloads, as well as the special demands of their intra-dependent components. In this paper, we present an offline mechanism for scheduling batches of brokered data-intensive workloads, which can be extended to an online setting. The objective of the mechanism is to decide on a packing of the workloads in a batch that minimizes the broker's incurred costs, Moreover, considering the brokered nature of such workloads, we define a payment model that provides incentives to these workloads to be scheduled as part of a batch, which we analyze theoretically. Finally, we evaluate the proposed scheduling algorithm, and exemplify the fairness of the payment model in practical settings via trace-based experiments.