Research On Price-driven Resource Allocation Scheme In Mobile Edge Computing

Mobile edge computing(MEC)nodes are physically closer to the terminal than to the Cloud Center,so users can enjoy cloud services by only one-step distance.However,MEC servers are usually equipped with very limited resources,so a helpful mechanism is needed to control user offloading to make the network feasible.In the MEC system,pricing is an important economic factor.It is very important to design a good pricing Scheme to maximize the revenue of cloud service operators and optimize the quality of services for users.The existing resource allocation mechanism based on the centralized method needs to obtain all mobile information to make resource allocation decisions,which will greatly increase the complexity of the algorithm.Many literatures that use game theory to implement computation offloading in MEC system,only consider one or two of the user's energy cost,delay cost,and currency cost when optimizing the user utility function,and rarely consider the combination of these three as a optimization target for computation offloading.Moreover,when constructing utility functions,some literatures did not specify the types of computing resources,or even analyzed specific function expressions.Most of the existing work on literatures ignores the user's service preferences,and different users may show different preferences when executing different types of applications or in different states.In addition,many literatures do not provide a comparative analysis between different pricing policies,because most research on edge computing only considers one pricing policy,so it is difficult to measure the advantages and disadvantages of various pricing policies.In view of this,this paper researched in two scenarios,a multi-user MEC system scenario where multiple mobile devices compete for the service of one MEC node and a multi-server MEC system scenario where multiple mobile devices compete for the service of multiple MEC nodes.This paper used game theory to design a price-based distributed resource allocation scheme to manage users' offloading tasks.The Stackelberg game models of single-leader and multiple followers and multiple-leader and multiple followers are formulated to describe the interactive between edge clouds and users.During the game,the edge clouds are the leaders,setting prices based on limited computing capability to maximize its revenue and limit users' offloading.For a given price,each user as a follower makes an offloading decision locally to minimize its own cost,which is defined as the weighted average sum of delay cost,energy cost and currency cost.The weighting factors can be adjusted to reflect the different preferences of users.This paper also proposed algorithms of resource allocation schemes with two pricing policies: Uniform pricing policy and Differentiated pricing policy,and then compared the performance of the two proposed schemes with other resource allocation schemes.The experimental results show that the two price-driven resource allocation schemes proposed in this paper can effectively increase the revenue of users and edge cloud in the system in the above two scenarios.In addition,this paper also compared the performance under different weighting factor ratios,and the impact of different user numbers and different MEC server computing capability on system performance.