Research On Cloud Resource Allocation Algorithm Based On Multi-Objective Optimization

As a new computing model, cloud computing has been applied to many fields, pooling service resources and providing various computing and storage services on-demand. Therefore, under the limited resources condition, designing resource scheduling and allocation schemes scientifically and optimally to balance cloud services capabilities and users'requirements have become an important research topic in the cloud computing.After analyzing distribution of cloud computing resources, the thesis studies two key issues of cloud computing resource management. The main contents are as follows:(1) Methods of resource abstraction among Service Providers (SPs) and Infrastructure Providers (IPs) now are unbalanced load and waste resources. So, the thesis proposes a new resource abstraction method, which utilizes Mixed-integer Linear Program (MILP) to integrate cloud resources. At the same time, the policy solves Linear Programming (LP) of cloud resources by a way of partial exhaustive. Compared with the current mainstream resource abstraction algorithms by experiments, the algorithm not only reduces the computing complexity to a certain extent, but also decreases by 8.2% wasting resources.(2) The existing cloud resource allocation strategies only take users'satisfaction SPs'profit as the optimization objective and have some limitations, the thesis designs a novel multi-objective optimization function, where the total time and total overhead to complete tasks are the influence factors and the optimization goals for balancing requirements from users and SPs. On the basis, the thesis presents a new Particle Swarm Optimization (PSO), which utilizes Damped Motion (DM) to periodically adjust inertia weight of the particle and disturbs the global optimal solution so that it realizes multi-objective optimization. Compared with the current mainstream resource optimization algorithms by experiments, in the case of the same tasks and resources, the algorithm reduces the total running time and total overhead to complete tasks.