The computing power network has become a crucial element in addressing the exponential growth in computing demands of future applications. The microservices model decomposes applications into loosely coupled, interdependent, fine-grained microservices. It is a significant challenge to efficiently utilize computing resources for creating optimal execution plans for dependent tasks in concurrent microservices architectures. To address this issue, this paper introduces container technology into the computing power network, establishing a container-based architecture and constructing a computational model to reduce average latency and network resource consumption while balancing resource load. Furthermore, this paper proposes a Feedback-Driven Bi-level Particle Swarm Genetic Programming algorithm (FD-PSGP). This algorithm integrates particle swarm and genetic algorithms within a bi-level planning model, featuring improved particle representation and update mechanisms based on problem-specific characteristics, and the design of effective individual representation and mutation strategies. Experiment results demonstrate that the FD-PSGP algorithm effectively reduces average latency, enhances resource and network load balance, and lowers network resource consumption.