The Research Of Cloud Datacenter Resources Scheduling Mechanism

In recent years, with the rapid development of Internet and virtualization technology, cloud computing, which providing users with on-demand services, has become a research hotspot. Under the environment of cloud computing, the datacenter, consisted by hardware and software, is a loosely coupled resource sharing architecture. The existing cloud computing’s inadequacies are as following three aspects:1. For lacking of real adequate and effective transaction of global bidirectional-way selection, the revenue of most of cloud resource provider is too low.2. Since not fully considering the scheduling of multi-dimensional cloud resources, existing cloud computing’s utilization for multi-dimensional cloud resource is too low.3. Because existing cloud datacenter does not fully consider the energy consumption of communication between the cloud tasks, its energy consumption is too high. In order to solve the above problems, this thesis proposes a novel resource scheduling mechanism inside the datacenter. This mechanism is based on the improved double auction bidding scheduling strategy which could satisfy most of the cloud market traders’trading needs. It also makes use of high efficient multi-dimensional cloud resource scheduling method which is integrated of heuristic particle swarm algorithm, genetic algorithm and ant colony algorithm. In addition, this mechanism achieves integrated scheduling scheme which could shut down idle processors under the premise of original system performance by the migration of virtual machines. Verified by a large number of experiments, this mechanism has advantages of cloud resource providers’ high yield, multi-dimensional cloud resources’ high utilization and cloud datacenters’ low power consumption. And the main research results are as follows:1. Bidding scheduling strategy based on double auction theory. Not only the existing cloud datacenter is not the real adequate and effective transaction of global bidirectional-way selection, but also most of the cloud market participants’trading requests have not been met. In addition, the revenue of cloud resources provider is too low and there is a serious problem of cloud market trading imbalance. For the limitation of transactions’unbalance in existing cloud market, this thesis analyzes the flexible and dynamic nature of cloud datacenters’bidding scheduling, and proposes a bidding scheduling strategy based on double auction mechanism, particle swarm optimal algorithm and genetic optimal algorithm This scheduling strategy could meet most of the cloud market traders’ trading needs, and improve cloud resource providers’ revenue. In this thesis, we verify the feasibility of bidding scheduling strategy based on double auction theory by CloudSim experimental architecture. Compared with the existing scheduling strategy, this scheduling strategy promotes cloud resources providers’ revenue growth rate to45%and trading volume growth rate to37.1%.2. Multi-dimensional cloud resource efficient scheduling method based on integer Markov theory. Not only the existing cloud datacenter is not the real multi-dimensional cloud resource scheduling, but also the resource utilization of multi-dimensional cloud resources (computing, storage and bandwidth) is too low. In addition, there is a serious problem of fragments resource waste existing in the cloud resources. For the limitation of the low multi-dimensional fragments resource utilization, this thesis summarizes the factors of affecting the multi-dimensional cloud resource scheduling method, reflects factors’ impact in form of parameters, and proposes a multi-dimensional cloud resource efficient scheduling method based on integer markov theory and combined with genetic algorithm and parallel ant colony algorithm. In addition, this scheduling method is truly multi-dimensional cloud resource scheduling and improves the utilization of multi-dimensional fragments resource. In this thesis, we verify the effectiveness of efficient scheduling method based on integer markov theory by CloudSim experimental architecture. Compared with the existing scheduling method, we find that this scheduling method promotes growth rate of multi-dimensional cloud resource utilization to50.31%.3. Low-power integrated scheduling scheme based on cloud tasks. With the expanding of the size of cloud datacenter, the energy consumption of communication between cloud tasks is continuously mushrooming. For the limitation of without considering the high energy consumption of communication between cloud tasks in existing research result, this thesis abstracts the inherent law of low-power scheduling scheme, and proposes a low-power integrated scheduling scheme combined with principles of resource integration, particle swarm optimization and tabu algorithm. In addition, this scheduling scheme also takes full account of the energy consumption of communication between cloud tasks, and reduces the energy consumption of the cloud datacenter. In this thesis, we verify the stability of low-power integrated scheduling scheme by CloudSim experimental architecture. By comparing with the existing scheduling scheme, we find that this scheduling scheme makes decline rate of energy consumption in cloud datacenter to60.81%.