Pricing the cloud: Resource allocations, fairness, and revenue

As more businesses use the cloud for their computing needs, datacenter operators are increasingly pressed to perform effective and fair allocation in this multi-resource, multi-tenant setting. The presence of multiple resources allows an operator to offer different types of pricing strategies (e.g., bundled vs. unbundled) that can have different effects on its revenue. Pricing also affects the demand and resource allocation decisions across clients who typically require different ratios of each resource (e.g., CPUs, memory, bandwidth) to process their jobs, which results in a complex trade-off between fairness and revenue maximization. We develop an analytical framework to investigate the fairness and revenue tradeoffs that arise in a datacenter's multi-resource setting and the impact of different pricing plans on the operator's objective. We derive analytical bounds on the operator's fairness-revenue tradeoff and compare tradeoff points for different pricing strategies on a data trace taken from a Google cluster.