An Energy-Efficient Metaheuristic for Offloading Security-Critical Tasks in Mobile Edge Computing with Integrated DPUs

In mobile edge computing (MEC) paradigm, for the security-critical computation tasks offloaded from the mobile device, the MEC server needs to decrypt the encrypted task data before task execution, leading to heavy computation load on the CPUs of the MEC server. The integration of data processing unit (DPU) into the MEC server can release CPUs from data decrpytion for the purpose of reducing computation and energy overheads. This paper studies a task offloading problem in a MEC system with integrated DPUs, which aims at reducing the total energy consumption under the deadline constraint on task completion time. We formulate the studied problem as an combinatorial optimization model and propose a group mapping-based cuckoo search (GMCS) metaheuristic algorithm to explore high-quality energy-efficient solutions to the formulated model. The proposed GMCS introduces a group mapping operator and a greedy-based task offloading strategy for converting cuckoo individuals into offloading solutions and evaluating the objective values of the converted solutions, respectively. We provide a theoretical analysis to justify that the mapping operator can improve the diversity of the converted solutions and in turn the metaheuristic’s searching capability. We create various testing instances for a MEC system equipped with DPUs to demonstrate the proposed GMCS produces high-quality solutions to the studied offloading problem, with reduced energy consumption compared with baseline algorithms.