Dynamic Service Caching Algorithm Based on D3QN in Satellite-Terrestrial Cooperative Edge Computing Networks
Jihong Zhao, Zhen Liu · 2024
With the development of the Internet of Things (IoT), IoT devices are generating an increasing number of service requests with low latency requirements.Deploying edge servers near the network edge close to IoT devices, and fully utilizing the limited resources of the servers to cache specific executable programs, databases, etc., can reduce the latency of service requests from the devices.However, due to the temporal variability of IoT device service requests and the dynamic nature of edge networks, selecting the optimal service caching strategy to reduce service request latency under limited resources (CPU resources, storage capacity) becomes more complex.To address this issue, a dynamic service caching algorithm based on the Dueling Doubling Deep Q Network (D3QN) is proposed, considering the satellite-terrestrial collaboration to provide services for IoT devices.In this algorithm, the device acts as an intelligent agent, independently deciding on caching modes and cache replacement decisions.First, the device autonomously selects caching nodes for service caching based on its current location.Secondly, a cache replacement action based on service cache value is designed within the algorithm to address situations where caching nodes have no available storage capacity.Experimental results show that compared with DDQN, NS-D3SC and FIFO algorithms, the proposed algorithm performs excellently in terms of service request latency and cache hit rate, reducing the average latency by 8.02% to 25.01% and increasing the cache hit rate by an average of 7.05% to 36.94%.