Energy Efficient Service Provisioning in Computing Power Network over OSU-Based OTN

With the development of 5G and edge computing, and the emergence of new applications, computing power network (CPN) is becoming an important solution for service provisioning. Optical service unit (OSU)-based optical transmission network (OTN) is introduced into CPN to provide high-quality transmission. Meanwhile, with the aim of carbon neutrality, energy efficiency is becoming a significant factor in CPN which actuates the CPN to take into account multiple operation objectives. However, the high dimensional computing and network resources and multiple objectives induce enormous computational complexity for service provisioning. This paper investigates the complexity reduction method for energy-efficient multiple services provisioning in the CPN over the OSU-based OTN. We first formalize the problem as a multi-objective quadratic integer programming (MO-IQP) to optimize the multiple objectives, including cost, load balancing, and carbon emission. When solving the MO-IQP problem, the whole request set is partitioned into subsets with spectrum clustering-based community detection, with requests in each subset being similar. Then, the constraint-aware network pruning is conducted for each request subset to shrink the solution space for optimization. The nondominated sorting genetic algorithm (NSGA-II) is adopted for each request subset with the corresponding pruned network. Simulation results demonstrate that convergence speed is accelerated and the performances are guaranteed.