Design And Implementation Of Task Resource Requirements Measuring System In Cloud Computing

Through the usage of virtualization technology to reorganize the underlying physical resources and virtualizing them, cloud computing is forming a huge pool of resources,computing center will provide these resources in a flexible and dynamic type of service to the user, which enable the users to obtain the required computing, networking, storage and other services. Resource provisioning technology is a concentrated expression of the flexibility of cloud computing, it has been an important technology to improve system availability.Currently, a variety of public cloud platform is flourishing. A variety of public cloud platform universal resource provisioning techniques aim to reduce data center power consumption, minimize cost and maximize profit. Even good support for public cloud platform, these techniques are not important in a dedicated private cloud. Each job runs in dedicated private cloud computing platforms will generally run repeatedly, each time differ in the input data set. Public cloud platform take commonality into consideration, it do not optimize this kind of platform feature. In addition, dedicated private cloud computing platform is used by professional users, it requires the user to describe task resource requirements by itself. But faced with vague requirements, users often greedily describe task resource requirements, resulting in wasted platform resources.For defects of the current dedicated private cloud computing platform task resource requirements description, this paper presents a description of the resource requirements in the form of data structure, designs and implements automated measure tasks resource requirements prototype system. The measuring system starts by allocating small amount of resources to the task. By using the virtual machine monitoring module determines if the task exists resource bottleneck, and according to the algorithm proposed, the system migrates the task to a virtual machine with more resources in. Aliquots proposed heuristic algorithms and dynamic programming algorithm results in a reasonable balance between the growth in the number of inter-virtual machine migration with the final remaining resources.Test results show that aliquots heuristic algorithms and dynamic programming algorithm are much better than the exponential growth of the slow start algorithm, providing a more rational resource growth model as a dedicated private cloud computing platform for virtual machine scheduling optimization provide new ideas.