Efficient and Fair Bandwidth Scheduling in Cloud Environments

Abstract

Hundreds of thousands of servers from data centers are operated to provide users with pay-as-yougo infrastructure as a service, platform as a service, and software as a service. Many different types of virtual machine (VM) instances hosted on these servers oftentimes need to efficiently communicate with data movement under current bandwidth capacity. This motivates providers to seek for a bandwidth scheduler to satisfy objectives, namely assuring the minimum bandwidth per VM for the guaranteed deadline and eliminating network congestion as much as possible. Based on some rigorous mathematical models, we formulated a cloud-based bandwidth scheduling algorithm which enables dynamic and fair bandwidth management by categorizing the total bandwidth into several categories and adjusting the allocated bandwidth limit per VM for both upstream and downstream traffics in real time. The simulation showed that paradigm was able to utilize the total assigned bandwidth more efficiently compared to algorithms such as bandwidth efficiency persistence proportional sharing (PPS), PPS, and PS at the network level.