Research And Implementation Of Distributed Multi-Agent Heterogeneous Scheduling Algorithm Based On Pi-calculus

Task scheduling is one of the core issues in the field of distributed computing.It can not only improve the computing power effectively,but also ensure that applications can run reasonably and effectively,directly affecting the overall performance of the distributed computing environment.Static heterogeneous scheduling is more universal,and it is also the most studied scheduling algorithm by scholars.Among the static heterogeneous scheduling algorithms,the heterogeneous earliest finish time(HEFT)algorithm is a classic algorithm with better scheduling length and robustness.Since the scheduling problem is NP-hard problem,how to improve the efficiency of the scheduling algorithm has always been one of the research directions of many scholars.Distributed computing systems can be divided into two types: isomorphic and heterogeneous.With the development of computer networks and computer systems,isomorphism is relative,and heterogeneity is absolute and inevitable.Considering that heterogeneous distributed computing system is absolute and more general,and the static scheduling algorithms are the most studied by scholars at present,this paper chooses static heterogeneous scheduling algorithms for research.Among many static heterogeneous scheduling algorithms,HEFT algorithm is a classical algorithm with good scheduling length and robustness.At present,most of the underlying models for task scheduling algorithm are serial models in nature.However,Pi-calculus is a mature parallel model that supports the transmission of channel names,which can easily realize the complex structural changes of communication systems.So it is very suitable for combination with distributed task scheduling.The cooperation,agency,intelligence,persistence and learnability of Agent-oriented technology make it very suitable for the implementation of distributed task scheduling.This paper combines Pi-calculus with multi-agent technology and applies it to the heterogeneous earliest finish time algorithm.We make use of the parallel computing characteristic of Pi-calculus,which also can use the channel to transmit information.Besides,combining with Multi-Agent technology can improve parallelism further.So we propose a distributed Multi-Agent model based on Pi-calculus which enables distributed resources to be used more efficiently.The main work is as follows:(1)Define the distributed Multi-Agent heterogeneous scheduling algorithm model based on Pi-calculus.And then model the HEFT algorithm in two stages,and give the channel graph and expression of processes.(2)Extend the Nomadic Pict language that inherits Pi-calculus and adds Agentrelated functions,supply its file processing functions.Then use the extended Nomadic Pict to program the model.(3)Construct DAG maps with different topologies for experimental comparison.In this paper,the Socket communication mechanism and multi-threading technology are used to assign the task to the HEFT algorithm implemented by multiple Clients.And then implement the HEFT algorithm based on the Pi-calculus without the multi-agent technology.At last,we implement the distributed Multi-Agent HEFT algorithm based on Pi-calculus and then compare and analyze the results of experiments with the two above.Through the experimental comparison results,the distributed Multi-Agent heterogeneous scheduling algorithm model based on Pi-calculus can improve the efficiency of the algorithm and make better use of distributed resources,which is of great significance to scheduling algorithms and related applications.