Scheduling Methods Of Cloud Computing For Simulation Based Optimization

The main goal of simulation optimization is to select the optimal design,and the another goal is to reduce the total number of simulation executions in the process of selecting the optimal solution from many solutions so that the efficiency of selecting the optimal solution can be improved.As higher of complexity and randomness with the practical problems,simulation optimization requires multiple independent testing of each scheme to obtain a more reliable optimal scheme,so the simulation optimization requires high and uncertain computing power.Cloud computing provides an economical,efficient,and flexible platform for simulation optimization.However,most of the current simulation optimization algorithms are for single-user execution,lacked of enough samples executed in parallel at each stage,resulted the unfully utilization of cloud computing resources and the unadaptability to cloud computing parallel,heterogeneous,shared and scalable features.Therefore,the research of cloud computing scheduling algorithm for simulation optimization problem,which can improve simulation optimization efficiency,reduce simulation optimization cost and meet multi-user Qo S requirements,has important theoretical and practical significance.In the light of the inability of the current simulation optimization method adapt to the parallel,heterogeneous,shared and scalable characteristics of cloud computing,the paper carries out research on key technologies such as scheduling methods of cloud computing for simulation based optimization,on the basis of in-depth analysis and research on traditional simulation optimization methods.Key work and innovations include:1)Lacked of sufficient parallel execution at each stage,traditional OCBA algorithms unable to fully utilize the parallelism,scalability,and low cost of cloud computing.To this end,a distributed asynchronous optimal computing budget allocation algorithm(DA-OCBA)algorithm for cloud computing is proposed.To maintain load balancing,the DA-OCBA algorithm improves the traditional OCBA algorithm by distributing the solution to each computing node for execution one by one.DA-OCBA improves the traditional OCBA algorithm by analyzing the staged results,pre-calculating,and using the results of the advance running,which can fully utilize the idle cloud computing resources in parallel.It shows by the experimental results that the simulation optimization efficiency of DA-OCBA is significantly higher than that of OCBA on the basis of the same simulation optimization results.As the computing resources increase,the speedup of DA-OCBA to the OCBA increases linearly.2)The traditional simulation optimization algorithms are all oriented to a single user task,and the computing resources of “cloud” can be shared by multiple users,resulting in computational node load imbalance when multiple users use cloud computing for simulation optimization.Other issues,it is unable to meet the multi-user quality of service(Qo S)requirements.To this end,a scheduling algorithm for cloud computing multi-user Qo S is proposed,which can sort dynamically based on user Qo S(user task workload and priority),and it can schedule computing resources non-preemptively according to the load balancing method of greedy algorithm.It shows by the experimental results that the proposed method can reduce the idle time of computing resources on the basis of multi-user Qo S and load balancing as much as possible,and effectively improve the computational efficiency of multi-user simulation optimization.As the research results shows above,a multi-user simulation optimization system based on cloud computing is designed and implemented,solving the unadaptability of the traditional simulation optimization method to the multi-user cloud computing environment.According to the comprehensive test,the system can provide an efficient method to solve the simulation optimization problem under cloud computing,it can fully and effectively schedule cloud computing resources,and the system can well adapt to the parallel,heterogeneous,shared and scalable characteristics of cloud computing.