The Research Of Dynamic Double-Level Layers VM Scheduler Strategies

The development of cloud computing is an inevitable product of information technology and time of big data. Cloud computing integrates grid computing, concurrency computing and virtual technology, being a new mainstream research direction. The increasing advancement bring great impact force to information age and its benefit is clear. Cloud computing provide a new thought:virtualization, pay-as-you-go, service. The application and infrastructures and mechanisms for the delivery underlying are separated by cloud, which increases collaboration, agility, expandability, usability and reduces the cost by optimized, effective computing.The thought of virtualization is the most important in the realization of cloud computing. However, the resources in cloud environment are not made full use. The traditional single task scheduling algorithms are still used in today’s cloud task scheduling models, mainly the best match scheduling strategy based on resource filter. But in the cloud environment, customers purchase the services they need and they want to complete as many as tasks with the limited storage resource and compute resource, which cannot be satisfied by traditional task scheduling models.A new virtual machine scheduling framework is proposed in the paper, allocating the resources in data center to users effectively. This scheduling strategy innovates from three aspects.1. Introducing the mechanism of two-layer queue. The first layer is admission queue, to control users’ access permission; the second one is processing queue. According to uses’ types, queuing time and the percentage of this sort of users to all users, determining the processing priority. Using FS-ALOHA algorithm to detect and manage collision, to guarantee all requests can get response in time. If users undo the requests, they can be cancelled in the queue, which can increase the efficiency and avoid repetitive operation.2. Introducing monitor to supervise the underlying virtual machine and physical servers. Because the user type and task type are different, the indexes monitors supervise are different. This paper propose seven sorts of monitor. This timely monitor can detect the statuses of different indexes in the system. In the same time, deploying ant colony search algorithm and by this method, the monitors can manage the processors’execute capability and processing capability. Users’requests can get introductions of monitors to the processors. 3. Introducing routing key to bind queue and monitor. Because the user type and task type are different, the indexes monitors supervise are different. After users submit their requests, the system attach tags (routing key) to different requests.Processing queue and monitor are established binding relationship by routing key creatively to a new effective and secure scheduler framework (QRM-framework). Through the theoretical analysis, this model can reduce error rate and fragment while increasing scheduler efficiency.