(Montana State University - Bozeman, College of Engineering, 2017) Yaw, Sean; Chairperson, Graduate Committee: Brendan Mumey
The performance of distributed applications is heavily dependent on the interplay between the applications and the underlying network. Disparity between the requirements of the applications and the capabilities of the network leads to degraded application performance, which in turn results in a drop in application usage or revenue. For example, many real-time interactive applications require lower latency than the public Internet provides, resulting in a poor experience for application users. At other times though, applications fail to effectively utilize all network capabilities. For example, conventional electrical appliances are currently unable to leverage the increased communication capabilities provided by the future smart power grid to decrease costs or modify consumption. Scheduling is an optimization technique to temporally and spatially allocate resources in such a way as to achieve some desired parameter optimization, such as minimized cost. In this dissertation, I study the use of scheduling techniques to counteract application performance degradation present due to the disparity between application requirements and network capabilities. I explore this disparity in both the smart grid and cloud networks, and propose novel algorithms that rely on numerous algorithmic techniques to realize application performance increases.
