Off-campus WSU users: To download campus access dissertations, please use the following link to log into our proxy server with your WSU access ID and password, then click the "Off-campus Download" button below.
Non-WSU users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Date of Award
Industrial and Manufacturing Engineering
Kai . Yang
Daniel . Grosu
Mobile Edge Computing (MEC) is the latest technology introduced with the aim of
reducing the response time of mobile applications by allowing them to perform their computation
at the edge of the network. Efficient resource allocation on edge servers is one
of the main challenges in edge computing. Due to the mobility of users, a poor allocation
might impose high execution costs. Computation offloading in edge computing has to
consider several issues that were not present in the data-center or cloud computing settings.
After the initial placement, mobile users may move to different locations. Therefore, an optimal
offloading decision made at the time of receiving a request may not remain optimal
for the whole duration of user’s application execution. In addition to this, the availability
of servers’ resources may change over time. Therefore, an efficient offloading algorithm
must be adaptive to this dynamic setting. When it comes to practice, there are lots of
uncertainties in the network that make optimal decision making demanding.
In this dissertation, we develop stochastic programing methods for resource management
in edge computing systems. We aim at developing novel models which consider
important nondeterministic parameters in edge computing systems, such as task arrivals,
mobility of users, and resource requirements of mobile applications. Another concentration
of this research is to design efficient solution algorithms which are able to give optimal
on near optimal solutions in a reasonable amount of time.
We address a very important problem in the management of MEC systems, that is, the
problem of finding an efficient application placement on the edge servers such that the cost
of execution is minimized. We develop a multi-stage stochastic programming model for
the application placement problem in edge computing systems and design a novel parallel
greedy algorithm based on the Sample Average Approximation (SAA) method to solve it.
In our experimental analysis, we use the random walk method to model the mobility of
users. We also model the problem of energy-aware application placement in edge computing
systems as a multi-stage stochastic program, where the objective is to maximize
the QoS of the system while taking into account the limited energy budget of the edge
servers. To solve the problem, we employ a parallel SAA algorithm, and conduct an extensive
experimental analysis to evaluate the performance of the proposed algorithm using
real-world trace data for the mobility of users. We also propose a risk-based optimization
method based on chance-constrained programming for application placement in MEC systems,
where the objective is to maximize the total QoS of the system. In this model the
resource requirements of applications is assumed to be a nondeterministic parameter. We
use SAA to solve the chance-constrained program, and develop a learning-based algorithm
to solve the SAA model.
Badri, Hossein, "Stochastic Optimization Methods For Resource Management In Edge Computing Systems" (2019). Wayne State University Dissertations. 2214.