MiceTrap: Scalable traffic engineering of datacenter mice flows using OpenFlow

Abstract

Datacenter network topologies are inherently built with enough redundancy to offer multiple paths between pairs of end hosts for increased flexibility and resilience. On top, traffic engineering (TE) methods are needed to utilize the abundance of bisection bandwidth efficiently. Previously proposed TE approaches differentiate between long-lived flows (elephant flows) and short-lived flows (mice flows), using dedicated traffic management techniques to handle elephant flows, while treating mice flows with baseline routing methods. We show through an example that such an approach can cause congestion to shortlived (but not necessarily less critical) flows. To overcome this, we propose MiceTrap, an OpenFlow-based TE approach targeting datacenter mice flows. MiceTrap employs scalability against the number of mice flows through flow aggregation, together with a software-configurable weighted routing algorithm that offers improved load balancing for mice flows. © 2013 IFIP.