Towards predictable datacenter networks

TLDR

In multi‑tenant datacenters, shared networking causes performance variability that degrades application performance, makes tenant costs unpredictable, and reduces provider revenue. This work argues for extending the tenant‑provider interface to explicitly expose network performance guarantees. The authors design virtual network abstractions that balance performance guarantees, tenant costs, and provider revenue, and implement them in the Oktopus system. Simulations and a 25‑node testbed show that virtual networks deliver more predictable tenant performance and can cut tenant costs by up to 74 % while keeping provider revenue neutral.

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, this paper makes the case for extending the tenant-provider interface to explicitly account for the network. We argue this can be achieved by providing tenants with a virtual network connecting their compute instances. To this effect, the key contribution of this paper is the design of 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 develop Oktopus, a system that implements the proposed abstractions. Using realistic, large-scale simulations and an Oktopus deployment on a 25-node two-tier testbed, we demonstrate that the use of virtual networks yields significantly better and more predictable tenant performance. Further, using a simple pricing model, we find that the our abstractions can reduce tenant costs by up to 74% while maintaining provider revenue neutrality.