Meet: Rack-Level Pooling Based Load Balancing in Datacenter Networks

Abstract

Datacenter networks enable multiple paths between hosts to provide large bisection bandwidth. It requires load balancers to cope with network uncertainties such as traffic dynamics and topology asymmetry. Existing edge-based load balancing schemes are usually faced with the problem of limited network visibility. This article proposes <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Meet</small> , a rack-level pooling based load-balancer deployed at the edge that can handle the aformentioned uncertainties. <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Meet</small> utilizes both passive information as well as active probing to comprehensively sense the network conditions with relatively low cost. <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Meet</small> dynamically reroutes flows effectively based on the visibility of the network condition. <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Meet</small> has been tested with extensive flow-level simulations against state-of-the-art load balancers. It outperforms Hermes by up to 10% in the experiments, and outperforms others solutions such as DRILL by up to 50%. <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Meet</small> requires no modifications to the switches and is feasible to deploy at the edge.