Research On Cloud Architecture And Scheduling Mechanism

With the rapid development of Grid computing, pervasive computing and computer communication technology, people increasingly want to bring resources, software and applications to better integrate and provide to users in a form of service, so the cloud came into being. In order to achieve the integration of resources and services, you need a more generic and service-oriented architecture clouds; the same time as the heterogeneous environments, distributed, autonomy and diversity of services of cloud computing, we put forward higher requirement on the cloud platform scheduling mechanism, so on the cloud structure and scheduling mechanism has got more and more attention by the industry.Based on the above point, this paper makes a in-depth study and discussion on cloud computing related technologies and existing cloud service architecture and scheduling mechanism, and summarizes the problems of current structure and operation mechanism of cloud. In order to solve these problems and take the characteristics of cloud computing into consideration, based on the research of fault-tolerant and scheduling in our laboratory, we design a generic model of cloud structure and flexible distributed scheduling mechanism, then make the complementation through hardware and software in our laboratory. The scheduling mechanism not only satisfies the user's Qos requirements of clouds, but also reduces communication costs of and adjusts the load balancing.For the resources or services under the cloud, our model has good scalability by using a distributed loosely coupled architecture to organize services. Because of the cloud's wide variety of service and its distributed architecture, we construct the service group to manage many different services. Service group not only manages different services effectively, but also makes the mapping from service request to the service type. Though service scheduling is a NP problem, this service scheduling model and the distributed algorithm overcomes the problems of centralized scheduling only by the central scheduler which is easy to cause bottlenecks. In this model, each node has an independent scheduler, which improves the service scheduling efficiency. Meanwhile, the scheduling algorithm considers a number of objects, including the earliest service completion time, costs and overhead, as well as load balancing in cloud, which considers scheduling optimization from the cloud users and cloud providers'point of view respectively, so it achieves the multi-objective and distributed service scheduling and satisfies various performance requirements by dynamic judgment rules and multi-objective scheduling.