Distributed Offloading Algorithm Design for Mobile Edge/Cloud Computing Systems
![restricted_to_institution](/assets/restricted_to_institution_icon-7d7fc9806cb362d0af51e67b7302f7f9dbd0e97c4946cdf9449c0a7bd69f8c7d.png)
Restricted (Penn State Only)
- Author:
- Qin, Xudong
- Graduate Program:
- Electrical Engineering (PHD)
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- February 20, 2023
- Committee Members:
- George Kesidis, Major Field Member
Bin Li, Chair & Dissertation Advisor
Ting He, Major Field Member
Uday Shanbhag, Outside Unit & Field Member
Madhavan Swaminathan, Program Head/Chair - Keywords:
- Mobile Edge/Cloud Computing
Distributed Algorithms
Wireless Networks
Distributed Offloading - Abstract:
- In mobile edge/cloud computing systems, it is important to have efficient offloading algorithms to help users reduce their computational costs by leveraging the advantages of powerful edge/cloud servers. Therefore in this dissertation, we study three different types of mobile edge/cloud computing systems and propose distributed offloading algorithms that minimize users' average task processing costs for each one of these systems. We take many factors (e.g., local processing latency, offloading latency, and energy consumption) into account, demonstrate the proposed algorithms' performance analytically, and perform simulations to validate our theoretical findings. We first study the large-scale mobile cloud computing systems and develop a distributed threshold-based offloading algorithm that allows each user to offload their task without knowing all other users' decisions. Then, we formulate the task process cost minimization problem as a symmetric game and characterize the sufficient and necessary conditions for the existence and uniqueness of the Nash Equilibrium (NE) of the system, assuming exponential service times. We further show that the proposed distributed offloading algorithm converges to the NE in finite iterations when the NE exists. We characterize the performance gap between cost under our proposed distributed algorithm and the minimum cost in terms of Price of Anarchy (PoA) when the cost of using cloud servers is high. Finally, we perform simulations to validate our theoretical findings and demonstrate the superior performance of our proposed algorithm compared to the well-studied probabilistic offloading algorithm. Then, we focus on the heterogeneous mobile edge computing systems, where all users have heterogeneous arrival rates, local service rates, offloading latency, and energy consumption when processing tasks. We develop a distributed threshold-based offloading algorithm based on edge servers' computational load, average offloading latency, average energy consumption, and edge server processing time, depending on the server utilization. We show that there always exists a unique Mean Field Nash Equilibrium (MFNE) in the large-system limit when the task processing times of mobile devices follow an exponential distribution, and the proposed distributed threshold-based offloading algorithm converges to the NE. Finally, we run simulations and show that the proposed distributed offloading algorithm performs well in both theoretical and practical simulation settings. Finally, we consider the user-level dynamic model in mobile edge computing systems. Each incoming user demands a heavy computation and leaves the system once its computing request is completed. We formulate an average energy consumption minimization problem and apply the classic stochastic network optimization frameworks to develop a Joint Distributed offloading and Wireless Scheduling algorithm that achieves the average energy consumption within O(1/K) of the minimum energy consumption required for network stability at the expense of total network workload growing with O(K), where K>0 is a system parameter. Finally, we show the performance of our proposed algorithm via simulations using both synthetic and real-world datasets, even in changing system statistics.