Achieving Application-Level Utility Max-Min Fairness of Bandwidth Allocation in Datacenter Networks

Abstract

Providing fair bandwidth allocation for applications is becoming increasingly compelling in cloud datacenters as different applications compete for shared datacenter network resources. Existing solutions mainly provide bandwidth guarantees for virtual machines (VMs) and achieve the fairness of VM bandwidth allocation. However, scant attention has been paid to application bandwidth guarantees for the fairness of application performance. In this paper, we introduce a rigorous definition of application-level utility max-min fairness, which guides us to develop a non-linear model to investigate the relationship between the fairness of application performance (utility) and the application bandwidth allocation. Based on Newton’s method, we further design a simple yet effective algorithm to solve this problem, and evaluate its effectiveness with extensive experiments using OpenFlow in Mininet virtual network environment. Evaluation results show that our algorithm can achieve utility max-min fair share of bandwidth allocation for applications in datacenter networks, yet with an acceptable computational overhead.