Research And Implementation Of Cluster Resource Management Technology

With the rapid development of the Internet, the World entered the age of the Internet. Because more and more individuals access to the Internet, the demand for computer services and applications are rapidly increasing. Followed by a rapid expansion of the computer scales. Under such a background, server cluster substitutes the single server to provide users with Internet services. Due to extremely rich Internet applications, and heterogeneous server trends increasingly significant, the pattern that a small cluster of traditional supports one or a small amount of application emergence a bottleneck: low resource utilization, high maintenance costs. In order to improve the overall utilization of cluster resources and reduce operating costs, the multiple clusters united in one cluster and resource sharing is an inevitable choice. However, how to achieve the unity of abstract heterogeneous cluster resources, how to use a variety of clusters with the right resources to achieve maximum utilization of resources is a key issue to achieve a cluster resource sharing platform. The researchers designed a lot of cluster resource management framework on this issue, the most prominent is the use of virtualization technology to mask hardware and software structure, realizing unified management of cluster resources. Based on this framework, many excellent virtualization resources management platform have been developed, such as Cloud Stack, Con Virt, Cloud Stack and Openstack and so on. However, these platforms still have some problems: 1, resource abstraction rough: abstracted resource dimension and supporting virtualization technology less; 2, management difficulties: these platforms face mostly for virtual machines and user configuration problems, and they do not support the mission of the life cycle management; 3, resource utilization is not high: There is no specific resource scheduling strategy, and these platforms can not describe the tasks of specific needs and heterogeneous attributes of the machine, impacting the task of efficiency.According to these resources management system defects, this paper puts forward a task-oriented supporting multiple applications cluster resource management framework, and achieve the appropriate cluster resource management system. The main contents and contributions are as follows:1) Construct a supporting multi-virtualized cluster resource management framework. Taking advantage of the benefits of virtualization, it eliminates coupling of application and physical clusters perfectly, making application development and cluster management completely independent. In addition, this framework supports a variety of virtualization technology and it is easy to expand. To achieve a multi-dimensional and fine-grained resource management strategies, The system manages not only CPU and memory resources, and also network I/O and other resources. This strategy can be further use of the unique advantages of virtual technology, a full range of system resources and fine-grained management of scheduling, to improve further resource utilization.2) The life-cycle management of cluster tasks incorporated into the cluster resource management framework, so that the cluster can quickly and efficiently allocate resources to the task and timely recovery of resources from the task. It avoids a series of conditions that the idle virtual machines increase the number of cluster load;3) To improve resource utilization, we achieve cheduling algorithms named DRBF.4) To achieve a multi-applications-oriented, multi-virtualization-supported technology resource management system Lcwyf. And go through actual application and the variety of experimental, it verifies that the system can significantly improve the cluster resource utilization, and make the cluster resource management more convenient and intuitive.