Parallel Component Composition and Performance Optimization Based on Agent Technology
Yunfeng Peng, Congming Shi, Guowei Gao, Jianan Wang, Hai Liu · 2021
In order to better assemble parallel component programs and optimize performance, different software agents are designed and used. The component connection agent is responsible for the component interface glue and data redistribution. The component execution agent and resource management agent cooperate with each other to deploy the components on the computing nodes that meet the resource requirements. Four different component adaptive strategies are defined. Different component adaptive agent, component execution agent and resource management agent cooperate with each other to complete the adaptive process of component and improve the performance of component. Resource management agent, load detection agent and component execution agent cooperate with each other to complete the load balancing work and improve the performance and throughput of the whole computing platform. Experiments on heterogeneous computer clusters demonstrate the effectiveness of the proposed parallel component assembly and performance optimization method based on agent technology. Compared with the traditional performance optimization methods, the method based on agent technology is flexible and has performance advantages.