Dynamic pricing for efficient workload colocation

Pricing models for virtualized (cloud) resources are meant to reflect the operational costs and profit margins for providers to deliver specific resources or services to customers subject to an underlying Service Level Agreements (SLAs). While the operational costs incurred by cloud providers are dynamic – they vary over time, depending on factors such as energy cost, cooling strategies, and overall utilization – the pricing models extended to customers are typically fixed – they are static over time and independent of aggregate demand. This disconnect between the cost incurred by a provider and the price paid by a customer results in an inefficient marketplace. In particular, it does not provide incentives for customers to express workload scheduling flexibilities that may benefit them as well as cloud providers. In this paper, we propose a new dynamic pricing model that aims to address this marketplace inefficiency by giving customers the opportunity and incentive to take advantage of any tolerances they may have regarding the scheduling of their workloads. We present the architecture and algorithmic blueprints of a framework for workload colocation, which provides customers with the ability to formally express workload scheduling flexibilities using Directed Acyclic Graphs (DAGs), optimizes the use of cloud resources to collocate clients’ workloads, and utilizes Shapley valuation to rationally – and thus fairly in a game-theoretic sense – attribute costs to customer workloads. In a thorough experimental evaluation we show the practical utility of our dynamic pricing mechanism and the efficacy of the resulting marketplace in terms of cost savings.