Optimize Algorithm for Increasing the Utilization Processor in mesh Multi-Computers Network
|Author(s)||Rahmat Zolfaghari, Meisam Musavi**|
|Issue Date||December, 2013|
|Publishing Date||December, 2013|
|Keywords||Utilization, multi-computers network, allocation processor, fragmentation, continuous, non continuous allocation algorithms|
Processor allocation is responsible for selecting a set of processors in order to run parallel work on them, while job schedule is responsible for determination of executing works. Job Schedule selects the next job for execution based on stated policy and then the processor allocation algorithm finds the free processors for the selected work. Until now, several continuous and non continuous techniques have been given for processor allocation in mesh multi-computers networks. Continuous allocation methods always try to allocate a free continuous sub-mesh with the same requested dimensional structure to the parallel input job. For this reason, it produces the internal fragmentation in the processors network. Non continuous allocation algorithms were produced with the aim of removing processors fragmentation. In non continuous allocation algorithms, message interference between different jobs and strugle to get communication resources increases network communication overheads due to the increase in path length passed by the message. This communication overhead is highly dependent on to the manner of free sub-meshes allocation and the manner of recording by the algorithm. This reasearch, a non continuous allocation algorithm called Efficient and Quick Non-Continuous Allocation (EQNA) algorithm has been presented for a two-dimensional mesh network with C programming language. The efficiency of this algorithm compared via simulator tool ProcSimity with other continuous and non continuous allocation algorithms(FF/BF, Paging and MBS). Simulation results indicate improved Utilization Processor in the given algorithm(EQNA).