Performance impact inference with failures in data center networks

Maintaining a data center network (DCN) is crucial to many services running on top of it, especially given its large scale with tens of thousands of network components. In this paper, we propose a method to infer performance change before failures really happen in data center networks, called Sibyl. Different from previous work, Sibyl relies on network topology information to infer network performance under failure scenarios without the overhead of active measurements. Specifically, we demonstrate that most important performance metrics can be obtained from two fundamental topological metrics. We develop efficient algorithms to obtain these two fundamental metrics, leveraging graph automorphism of various DCN topologies.