Research On Key Issues Of Cloud Computing For Performance Analysis Of Electrical Equipment

Manufacturing industry of the electrical and even traditional equipment are mostly dependent on the "experience",so there are several common problems of the low rate of accepted product、serious waste of resources and the long design cycle.In the Internet Era,"Made in China,2025"is a very important guide direction for reform and upgrade on traditional manufacturing industry and an intelligent manufacturing design method relating to internet、big data and cloud computing.In respect of pervasive computing,the traditional high performance computing has been analy sized with problems of high cost,difficult maintenance and long waiting time for jobs,because realizing intelligent manufacturing of electrical equipment is based on high performance computing and faced with problem of large-scale computation.To solve the problems above,cloud computing with the merits of pay-as-you-go、flexible and easy to take and dynamic extension has been introduced,a idea that cloud computing and finite element simulation system in the electromagnetic field have been integrated was provided and a simulation system finite element method&cloud computing was built.VMware vSphere,at first,was used for integrating of 5 server resources to form a virtual resource pool in which a kind of elastic cluster with 9 computing nodes for parallel computing was set up.After deeply analyzing cloud computing,an effective test scheme for performance of cloud computing is presented.The performance test on virtual machines and physical computers with same resource configuration has been conducted from aspects of CPU,memory,network,disk,the capability of cluster computing and the overall evaluation.The result shows that virtual resources in cloud computing has some performance penalties compared with physical resources.Management mechanism of cloud computing was given inclouding creation process process and ways to migration and backup in virtual resources.A successful migration of Parallel computing cluster was made form workstation to servers of Laboratory,which indicates flexibility of migration and backup in virtual resources better than traditional high performance computing.Secondly,comparison and analysis for MapReduce、MPI and OpenMpi of different parallel computing architectures about cloud computing have been made.OpenMpi suiting for large-scale intensive computing has been chosen of which message passing mechanism has been presented.Induction motor,transformer and the international benchmark problem of TEAM 7 in the electromagnetic field are chosen as the calculation cases of which calculation programs of static magnetic field are written based on Fortran.The parallel computing cases,with hundreds of large-scale computing nodes,have been carried out,and the distribution graph of computational time and rate under different computational scales and different nodes were painted.The task scheduling and fault tolerance mechanism of the multi computation nodes are introduced.For the imbalance problem of tasks with large scale in many computation nodes,an effective solution to computing task scheduling and fault tolerance mechanism were offeredAt last,the research proved that the parallelization of the large-scale finite element simulation of typical electrical equipment can be achieved in the elastic computing cluster which is deployed in the cloud computing platform.In contrast to parallel computing method for the traditional high performance computing,the elastic computing cluster in cloud computing has a series of merits,such as invested cost low,dynamic distribution,obtain by demand.Thus a flexible and convenient new high performance computing scheme is provided for upgrading from pure traditional manufacturing industry to intelligent manufacturing industry which is impelled by innovation.