Abstract

The shared nature of the network in today’s multi-tenant datacenters implies that network performance for tenants can vary significantly. This applies to both production datacenters and cloud environments. Network performance variability hurts application performance which makes tenant costs unpredictable and causes provider revenue loss. Motivated by these factors, in this talk I will make the case for extending the tenant-provider interface to explicitly account for the network. This can be achieved by providing tenants with a virtual network connecting their compute instances. To this effect, I will describe virtual network abstractions that capture the trade-off between the performance guarantees offered to tenants, their costs and the provider revenue.

To illustrate the feasibility of virtual networks, we have developed Oktopus, a system that implements the proposed abstractions. I will present the Oktopus design and briefly discuss evaluation results showing that our abstractions enable a symbiotic tenant-provider relationship with significant benefits for both entities.