Service Degradation Considering Computing, Communication, and Storage With Resource-Time Dimension in Computing Power Network

Service degradation has been proven to be an effective solution in network congestion scenarios, deploying as many services as possible at the cost of degrading service performance. Most existing service degradation schemes consider only computing and communication resources independently, without considering storage resources. However, storage resource is required to store the intensive computing data in the case of lacking computing resources. Thus, it is desired to design a collaborative service degradation strategy that jointly considers computing, communication, and storage resources, ensuring efficient service deployment and mitigating network congestion in CPNs. To address the above challenge, we formulate an integer linear programming (ILP) model that enables collaborative degradation in resource-time dimension, integrating computing, communication, and storage resources to achieve efficient resource management in a small network. To solve the same problem for large-scale networks, a deep reinforcement learning algorithm, DeepSD, is proposed to collaboratively optimize computing, communication, and storage resources by selecting degradation services and coefficients. Simulation results show that ILP and DeepSD reduce the blocking probability by 23% and 17% and improve computing, bandwidth, and storage resource utilization compared to the benchmark.