Ad Hoc Task Cloud Decision Unloading

With the development of mobile communication technology and internet technology, the mobile users hope that mobile terminals have the same performance as high-performance computers. But compared with the traditional computers, limited resources, short battery life and less memory have limited the performance of mobile terminals, so the computation-intensive tasks spend more time and energy on the devices with limited resources. By reasonable and effective computing task partitioning, it can not only make full use of the scattered resources in the network, but also improve the data processing capacity of the mobile terminals.In the Ad hoc cloud environment, there are two solutions for computing tasks offloading:(1) Divide the computation-intensive tasks into "pieces". That is to say, divide a task into multiple subtasks which have a small amount of computation, then distribute these subtasks to a number of mobile terminals for completion.(2) Directly assign a whole task to a HPC agent node for completion. The first solution is the problem of task partitioning. Based on the maximum flow minimum cut2-partition algorithm, this paper analyzes the multiple partition of related tasks under the multi-offloading model. The second solution is the problem of task scheduling. The second solution is the problem of task scheduling. According to the characteristics of the mobile ad hoc environment, this paper proposes simple additive weight(SAW) strategy, which considers energy consumption of the link and makespan in the the task scheduling process, and compare with the MH, MET, MCT, Min-Min, SAW strategy. The experiments show that the SWA strategy better balance energy consumption of the link and task completion efficiency.