Research On Elastic Resources Scheduling Technology In Cloud Environment

Elastic resources scheduling technology has become a basic problem of researching on cloud computing system, also to be the critical issue for building elastic cloud computing system. It is essential for cloud computing service providers to maximize resource utilization, save resources use cost and improve the efficiency of users’ applications. This paper concerns about cloud elastic resources scheduling based on a typical distributed cloud computing system of hierarchical data centers, and make an in-depth study of two issues of cloud resources allocation and dynamic resources scheduling. Specifically, the main contributions of this paper are as follows:(1) Cloud resources allocation method based on multi-user and single data centerTo solve the cloud resources allocation problem of multi-user and single data center, existed methods usually only focus on looking for the current local optimal solution and are lack of the consideration about global optimal solution, resulting in low comprehensive utilization of resources and significant performing differences among users’ virtual resources, so we propose a layered progressive resources allocation algorithm based on the Multiple Knapsack Problem(LP-MKP). Experimental results show that, LP-MKP algorithm is superior to the maximum idle resources based on the greedy algorithm and the heuristic algorithm based on the best sub-tree, and suits for cloud resources allocation in multi-user environment.(2) Cloud resources allocation method based on multi-user and multiple data centersTo solve the cloud resources allocation problem of multi-user and multiple data centers, existed methods are usually lack of the consideration that multi-user resources allocation can be performed simultaneously and users’ locations can impact on the service quality in multiple data centers(MDC) environment, resulting in low comprehensive utilization of resources and significant performing differences among users’ virtual resources, so we propose a resources allocation algorithm based on greedy algorithm in MDC environment(RAGA-MDC) and resources allocation algorithm based on multistage decision in MDC Environment(RAMD-MDC). Experimental results show that, RAGA-MDC and RAMD-MDC algorithms can significantly get better results which to be the sum of all service distances compared to random algorithm, RAMD-MDC algorithm can get excellent results compared to RAGA-MDC algorithm, but the computing efficiency is relatively lower than RAGA-MDC algorithm.(3) Dynamic resources scheduling technology in cloud computing systemTo solve the dynamic resources scheduling problem in cloud computing system, we divide the dynamic scheduling technology into two parts, elastically increase or decrease cloud resources and dynamically optimize the structure of resources. To solve the problem of determining the operation objects while elastically increasing or decreasing resources, we put forward a dynamic expansion algorithm of cloud resources based on dynamic programming(DEA-DP) aimed to minimize the network diameter of resources. To solve the problem of determining the migration objects while dynamically optimizing the structure of resources, we propose a dynamic optimization algorithm of cloud resources based on heuristic function(DOA-HF). Experimental results show that, DEA-DP algorithm has well improved the performance of virtual resources, DOA-HF algorithm can achieve the purpose of effectively and rapidly optimize the structure of resources, so as to enhance the operating efficiency of users’ virtual cluster.