Analysis And Research Of Resource Scheduling Management On Load Balancing In Cloud Computing

Task scheduling is an NP-hard optimization problem for cloud computing research, and load balancing is one of the hot researches of task scheduling. Load balancing technology enables each resource to work together to provide powerful computing capabilities. It plays an important role in balancing load, improving utilization of resource and the users’ quality of service, reducing response time of jobs and guaranteeing the Service-Level Agreement. Many domestic and foreign researchers have done a lot of researches related to load balancing technologies in cloud computing. They often use different load balancing strategies for different cloud systems to achieve optimal performance. This paper deeply analyses the advantages and disadvantages of the existing load balancing strategies, and respectively studies centralized load balancing strategy based on the classification of virtual machines and hierarchical hybrid load balancing strategy according to the different characteristics of small and large cloud systems.When the centralized load balancing strategy is applied to cloud computing system, the central scheduler searches target virtual machine for the task from the global scope, which may lead to low efficiency. And some strategies do not take into account the tasks’ priority which leads to low quality of service for users. Hence this research proposes a centralized load balancing method based on virtual machine classification. This method classifies all the virtual machines into three categories by using Na?ve Bayes algorithm. This makes the solution space clear and small and helps to improve the search efficiency. This method establishes a load balancing model by simulating the behavior of honey bees. By migrating the tasks from heavy load virtual machines to light load virtual machines, the number of migrated tasks is reduced and the cloud system can be balanced faster. Considering the task priority and choosing the target virtual machine with the least number of higher priority tasks than the migrated task makes the migrated one be handled earlier. This improves the quality of service for users.The hierarchical hybrid load balancing strategy rarely considers the communication quality, but it has effect on load balancing. And one scheduler is often set for one cluster which may lead to the scheduler with heavy workload. Sometimes the Service-Level Agreement violation rate is not low. To solve these problems, this paper proposes a hybrid load balancing strategy based on Multi-layer architecture and dynamic weighted method. All virtual machines are divided into clusters, and clusters directly connected are treated as neighbors. Tasks are scheduled among all the neighbor clusters at cluster-level, this helps reduce the overhead leaded by broadcasting information globally. The proposed method establishes a load balancing model between clusters based on the lowest communication cost by considering the communication quality. Hence the response time of tasks can be reduced.Two schedulers are set for each cluster and responsible for inter-cluster and intra-cluster tasks scheduling respectively which reduces the workload of one. Virtual machines in cluster are dynamically weighted by neural network method according to Service-Level Agreement.The availability of resources can be better reflected and better decision of scheduling task canbe made. And Service-Level Agreement violation rate is reduced by configuring spare virtual machines.