High Reliability Resource Scheduling Mechanism For Electromagnetic Computing Tasks

Electromagnetic computing tasks have the characteristics of large scale,high temporal and spatial complexity,and high computational precision requirements.Not only the demand for computing resources is large,but also have high requirements for the performance and reliability of resources.Traditional electromagnetic computing center mainly uses physical cluster to provide resources,which is limited by hardware equipment,cluster maintenance and application deployment cost is high,and the computing environment is single,can not provide different computing environment for different tasks.Secondly,traditional physical cluster can only dispatch and use the independent physical host,can not refine and isolate the computing,storage and network resources inside the host,therefore,in the multi-dimensional resource allocation scenario,it is difficult to realize the flexible resource segmentation according to the resource demand of the task,the utilization rate of the host resource is not high and the load balancing is coarse.Thirdly,when the host fails,can not achieve the online migration for task,easily lead to task failures and data loss,making disaster tolerance is poor,reducing the reliability.Cloud computing technology combines the advantages of parallel computing,distributed computing,and high performance computing,using virtualization technology to integrate a large number of heterogeneous resources,the user can allocate resources according to the needs of the tasks,having the flexibility of the computing environment configuration.Breaking away from the constraints of traditional physical equipments,improving the resource utilization,ensuring the calculation efficiency and reducing construction costs of the platform,making up for the physical cluster disadvantages,making it one of the first choice for high-performance computing architectures.The core of cloud computing is the abstraction and redistribution of resources,therefore,resource scheduling becomes a key problem in cloud platform.In the process of resource scheduling,we need to consider the factors such as resource distribution,resource availability,resource reliability,resource utilization and load balance,so as to provide high efficiency,high precision and high availability resource scheduling service for complex electromagnetic computing tasks.The main consideration of the traditional scheduling algorithm is whether the remaining resources can meet the task needs,without considering the availability and reliability of resources.In consideration of the high reliability requirement of electromagnetic computing task,aiming at improving the success rate of task execution,this thesis designs a resource scheduling mechanism(HRLB-ECT)with high reliability and load balancing for electromagnetic computing tasks.The main work is as follows:(1)Based on the fuzzy comprehensive evaluation method,taking the resource failure rate,MTTR,the success rate of the task execution and the available virtual service ratio of the resources as the evaluation indexes,the reliability of the physical resources is evaluated before the scheduling.According to the evaluation value,the reliable level of the resources is divided.The resources those meet the task reliable level constraints will be added to the scheduling resource pool,and the others will be removed,so as to protect the reliability of task runtime environment.(2)Taking the DRF algorithm as the prototype,this thesis puts forward the ECT-DRF algorithm for the electromagnetic computing task,which avoids the imbalance of the load between the resources,and ensures the calculation performance of the resources.The experimental results show that the HRLB-ECT scheduling mechanism can schedule the task to the relatively reliable resources and achieve the load balancing among the resources,which ensures the calculation performance and reliability,and improves the computing efficiency and task execution success rate.