Access to this thesis is limited to Boise State University students and employees or persons using Boise State University facilities.

Off-campus Boise State University users: To download Boise State University access-only theses/dissertations, please select the "Off-Campus Download" button and enter your Boise State username and password when prompted.

Publication Date

5-2019

Date of Final Oral Examination (Defense)

3-4-2019

Type of Culminating Activity

Thesis - Boise State University Access Only

Degree Title

Master of Science in Computer Science

Department

Computer Science

Major Advisor

Yantian Hou, Ph.D.

Advisor

Dianxiang Xu, Ph.D.

Advisor

Min Long, Ph.D.

Abstract

Traditional state-of-the-art cloud-computing is intrinsically suitable for processing computation-intensive image processing tasks due to its abundant computing and storage resource. However, the current cloud-based architecture requires remote data offloading, resulting in large transmission delay, making it highly unsuitable for current IoT based applications that are latency-sensitive. Our solution is to use the edge-computing technique to assist the cloud platform by processing the tasks on the edge servers that are placed close to the data sources. In this thesis, we apply edge-computing for two closely-related image processing tasks that are very common in modern IoT applications.

First, we consider a large-scale camera surveillance system for human re-identification. Our proposed solution considers the interplay between edge and cloud server, which we model mathematically and formulate as a bi-criteria optimization problem. We develop a measurement-driven offline algorithm that models the system's heterogeneity in terms of workload complexity. We propose an online algorithm which is based on linearization and a pipage rounding technique that efficiently solves the formulated optimization problem. The solution gives the decision variables that maintain the user-defined trade-off between system utility and responsiveness. In the second work, we develop a novel edge-assisted hierarchical platform. Our platform utilizes the processing capability at the edge to pre-process the workload to approximate the most relevant features, which are then precisely analyzed at the cloud. Considering various responding and placement strategies, we formulate a latency-minimization problem. Using a queuing mechanism, we capture the background workload's influence at the edge server and develop a two-stage algorithm to solve the problem. The solution gives optimal responding and placement strategies to process the body-detection task in the most time-efficient way. We implement our entire system on a commercial cloud platform and use extensive simulations and experiments to validate its efficacy and responsiveness in practice.

DOI

10.18122/td/1510/boisestate

Share

COinS