Zoom Meeting: https://umass-amherst.zoom.us/j/95544182236 Meeting ID: 955 4418 2236
Abstract
The past few years have witnessed significant growth in the use of distributed network analytics involving agile code, data and computational resources. In many such networked systems, for example, Internet of Things (IoT), a large number of smart devices, sensors, processing and storage resources are widely distributed in a geographic region. These devices and resources distributed over a physical space are collectively called a distributed service network. Efficient resource allocation in such high performance service networks remains one of the most critical problems. In this thesis, we model and optimize the allocation of resources in a distributed service network. This thesis contributes to two different types of service networks: caching, and spatial networks; and develops new techniques that optimize the overall performance of these services.
First, we propose a new method to compute an upper bound on hit probability for all non-anticipative caching policies in a distributed caching system. We find our bound to be tighter than state-of-the-art upper bounds for a variety of content request arrival processes. We then develop a utility based framework for content placement in a cache network for efficient and fair allocation of caching resources.
We develop provably optimal distributed algorithms that operate at each network cache to maximize the overall network utility. Next, we develop and evaluate assignment policies that allocate resources to users with a goal to minimize the expected distance traveled by a user request, where both resources and users are located on a line. Lastly, we design and evaluate resource proximity aware user-request allocation policies with a goal to reduce the implementation cost associated with moving a request/job to/from its allocated resource while balancing the number of requests allocated to a resource. Depending on the topology, our proposed policies achieve a 8% - 99% decrease in implementation cost as compared to the state-of-the-art.
Advisor: Don Towsley