Abstract

The adoption of cloud computing is hindered by a disconnect between tenant requirements and provider offerings. Tenants are primarily interested in predictable performance and bounded costs for their applications while cloud providers offer a resource-level interface. Consequently, tenants have to translate application requirements to specific resource demands, such as the number of virtual machines. We have been studying this disconnect in the context of data analytics applications. In this talk, I will present Bazaar, a system that enables predictable performance for such applications in multi-resource, multi-tenant data centers.

Bazaar allows tenants to express high-level goals and predicts the resources needed to achieve them. Since multiple resource combinations may achieve the same goal, it chooses the combination most suitable for the provider. Bazaar also enables a departure from today’s resource-based pricing to a novel job-based pricing model. I will show how Bazaar, by bridging the tenant-provider gap, can benefit them both.