Using Random Neural Network for Load Balancing in Data Centers

A data center which consists of thousands of connected computer servers can be considered as a shared resource of processing capacity (CPU), memory, and disk space etc. The jobs arriving at the cloud data center are distributed to different servers via different paths. In addition, the internal traffic between servers inside the data center needs to be load balanced to multiple paths between them as well. How to select the underutilized or idle paths for the traffic so as to achieve load balancing and throughput optimality is a big challenge. The Random Neural Network (RNN) is a recurrent neural network in which neurons interact with each other by exchanging excitatory and inhibitory spiking signals. The stochastic excitatory and inhibitory interactions in the network makes the RNN an excellent modeling tool for various interacting entities. It has been applied in a number of applications such as optimization, communication systems, simulation pattern recognition and classification. In this paper, we propose to use Random Neural Network (RNN) to solve the load balancing problem in data centers. RNN is able to achieve adaptive load balancing based on the online measurements of path congestion gathered from the network.