Mapping Parallel Workload onto Network Processors Using Simulated Annealing

Network Processors (NPs) are promising components to build a performance-scalable and function-flexible network systems. A clear trend has been observed that more and more NPs employ multiple simple processors to run multiple packet processing applications in parallel. The key challenge for application developers is how to program NPs for high performance. This paper presents an automated task scheduling technique based on Simulated Annealing(SA). By incorporating tasks dependency into scheduling list, SA can quickly remove the invalid mappings and evolve to the high quality solutions. Particularly, a transition probability is defined to help converge, which is depending on system throughput and control parameter ‘temperature’. Our mapping technique is able to take advantage of task-level and application-level parallelism to maximize system performance. The simulation results show that this proposed technique can generate high quality mapping comparing to other heuristics by mapping some sample network applications.