A Study Of Swarm Intelligence Optimization In Cloud Computing Resource Load Balancing

Technology innovation platform provides a new model for turning science technologyinto social productivity. As the information quantity on technology innovation platformcontinuously increase, artificial methods now cannot satisfy the realistic need. Therefore,new technology should be developed to meet the new requirements. Cloud computing ishelpful for solving this problem, but cloud computing has flowing defects, such as theproblems of usability unreliability, standardization, safety, load balancing, etc. To constructthe technical innovation platform, resource load balancing problems need to be solvedurgently. This paper sets a cloud computing resources load balancing model, in view of thecharacteristics of cloud computing resource node, in order to improve the resource loadbalancing degree and resources utilization rate.First, this paper analyses cloud computing needs on the technology innovationplatform, and describes in detail the characteristics of cloud computing resources(multi-source and heterogeneous, vast and dynamic changing resources, lower requirementsfor the servers, etc). The complex network has nonlinear interaction, is constituted by alarge element as well as dynamic, open, distributive, and is showing in swarm emergence.It is just coinciding with the characteristics of cloud computing resource nodes network.Therefore, cloud computing resources network nodes can be described by the complexnetwork, also has its characteristics, namely small world, scale-free, distribution of the nodedegree obeying power law distribution. The ability of resource nodes reflect by CPU rate,CPU utilization, memory size, memory surplus size, and hard drive surplus size. Resourceload balancing model aims at getting minimum load balancing degree and the shortest taskexecution time.In addition, through analyzing and comparing the characteristics of complex networkand particle swarm algorithm, it is demonstrated that they fit each other. So, although thereare many methods to solve the resource load balance model, but particle swarm algorithm ismore appropriate. Then this paper combines the particles updated characteristics in theparticle swarm algorithm with complex network characteristics of resources load balancemodel, elaborates a new particle swarm algorithm, and simulates it.At last, through comparing the improved particle swarm algorithm in this paper withthe LB-GA, LB-HA algorithm, the algorithm achieved in this paper is more suitable fordealing with cloud computing resources load balancing problem. And the population sizeand accelerated coefficient of the improved particle swarm algorithm is analyzed. The population size and the node number are relevant. The accelerated coefficient relates toparticles’ own experience and other particles’ experience, and if experience from their ownhas a higher influence, the corresponding accelerated coefficient is relatively larger. Settingvalue strategy and random strategy can be used to determine the accelerated coefficient.Contrasting the two strategies and it can be concluded that the random strategy is moreadvantageous for the particles to jump out of the local optimum.In this paper, the complex network theory is used to set up cloud computing loadbalancing model, and by using particle swarm algorithm to simulate and analyze, it can beconfirmed that this method has got better load balancing degree, task completion rate andcustomer satisfaction on cloud computing platforms.