Paper Info
Title
Opportunistic Coded Distributed Computing: An Evolutionary Game Approach
Abstract
Task offloading has been proposed and studied to overcome the problem of energy and computation constrained terminals. Computationally intensive tasks are often parallelable, and therefore the execution time can be further improved via a coded distributed computing (CDC) approach, as CDC offers robustness against stragglers by introducing redundant computational tasks. In this paper, we study a user-centric task offloading problem, in which the edge performs the offloaded computation with CDC. Furthermore, the extent of the straggler's effect on servers is also unknown to the user. This requires users to explore server and code settings of the CDC, and "opportunistically" select the best combo to maximize the utility. For simplicity, we refer to this scenario as opportunistic coded distributed computing. We formulate the problem as an evolutionary game in which each user is self-interested. The payoff is calculated based on the monetary cost of CDC-as-a-Service and total delay, weighted by user-defined parameter values. For the game solution, an evolutionary stable equilibrium (ESS) is used, i.e., probabilistic joint selection of server and code configuration. To obtain the ESS, we present an iterative algorithm based on the revision protocol. A theoretical analysis of equilibrium in terms of existence,uniqueness, stationarity, and stability is provided. Numerical simulations are conducted to support the theoretical findings and the adaption of equilibrium states to the hyper-parameters.
Year
DOI
Venue
2021
10.1109/IWCMC51323.2021.9498947
IWCMC 2021: 2021 17TH INTERNATIONAL WIRELESS COMMUNICATIONS & MOBILE COMPUTING CONFERENCE (IWCMC)
Keywords
DocType
ISSN
mobile edge computing, task offloading, computation offloading, coded distributed computing, evolutionary game theory
Conference
2376-6492
Citations 
PageRank 
References 
0
0.34
0
Authors
4
Name
Order
Citations
PageRank
Han Yu100.34
Niyato Dusit29486547.06
Cyril Leung389962.23
Dong In Kim43784220.90