Load balancing in data center networks with folded-Clos architectures

Data centers need to support a large number of dynamic resources such as virtual machines and storage volumes from any physical location within the data center. This has led to highly volatile and unpredictable traffic patterns. The emergence of applications that exchange large volumes of information have created large persistent flows that need to coexist with other traffic flows. Current load balancing protocols that rely on static hashing over equal cost paths do not take into consideration traffic volume nor network state. These protocols are no longer sufficient for achieving high bandwidth utilization. In this work we investigate load balancing in data center networks with folded-Clos architectures. We propose two load-balancing methods; “Selective Randomized Load Balancing” and “FlowFit”, the first method is an enhanced form of randomized load balancing, while the second is a form of dynamic load balancing that relies on a global view of the network state, for assigning flows to links. We build our topology using a centralized OpenFlow controller and open virtual switches. Our experimental and simulation results show a significant increase in bisection bandwidth over static hashing. These gains are achieved with minimal changes to the control and data planes.