Load balancing scheduling algorithm based on edge clustering in heterogenous systems

In order to give full play to the high parallel and high efficiency characteristics of heterogeneous signal processing platforms, relevant scholars at home and abroad have proposed many task scheduling algorithms with their own advantages to improve processing efficiency. However, for communication intensive tasks, the existing task deployment methods are no longer applicable because of their higher communication requirements and communication delay becomes the key to limit the performance of the platform. For communication-intensive signal tasks, this paper proposes a load balancing scheduling method based on edge clustering. Tasks are abstracted as directed acyclic graph, we aggregate the qualified task edges according to the rating function to reduce the computational complexity, and task duplication is used to replicate the tasks on the critical path to reduce communication overhead and dynamically balance processor load. Through the experimental verification of random task graphs and radar task, it is concluded that the algorithm has a good scheduling result for communication intensive tasks, which can achieve good processor load balancing and system scheduling length reduction.