Two-phase sample average approximation for video distribution strategy of edge computing in heterogeneous network

Abstract

Cooperative edge computing provides a good platform for edge storage and computing at the same time. When a single edge server cannot efficiently provide video services, multiple edge servers are needed to share resources for collaborative storage and computing. Existing collaboration schemes often ignore the coupling between video caching and distribution, resulting in inefficient caching strategies. Therefore, for the scenario where multiple edge servers provide video services cooperatively, the content access delay in the video service is minimized in this paper based on the video caching strategy on the slow time scale and the video distribution strategy on the fast time scale. The problem is modeled as a random integer linear programming problem on dual time scales. Moreover, an algorithm is proposed based on the sample average approximation technique. The algorithm first designs the video caching strategy of the edge server based on the statistical information of the arrival of video requests on the slow time scale and the expected video distribution strategy. Then, the video distribution strategy is optimized based on the designed caching strategy and the video request situation on fast time scales. Finally, through the cooperation of the two time scales, the content access delay is minimized, and the simulation results verify the advantages of the proposed scheme in reducing content access delay and improving storage hit rate. The storage hit rate is increased by 7.6%; the total content access delay is increased by 17.42%; the number of edge servers and the arrival rate of video requests are ahead of other methods by 5%; and the transcoding time is shortened by 4.56 s. It fully verifies that the design of a more efficient multi-user computing offload strategy in this paper can solve the problems of computing power, bandwidth, delay Energy consumption and other bottlenecks are of practical significance.