Research On Technology Of Virtual Machines Dynamic Migration Based On Based On Time Series Forecasting

With the rapid development of Internet technology and network technology,network users appears explosive growth, which has put forward faster, higherrequirements for server computing and data processing power. Therefore the Cloudtechnology arises for these needs. Cloud computing environment resources arevirtualized and highly flexible, and in addition to monitor resources of the physicalserver, but also need to run the virtual machine monitor. Cloud computing is businesscomputing model, in which users need to be reasonable billing. Existing resourcemonitor systems have not provided fine monitor granularity, and can’t get into theprocess-level information, can’t track the user to perform tasks in real time process,such as CPU, memory, and storage resource consumption. Cloud computing platformapplication and user needs are diversity and server node resources are heterogeneouslead to a system of nodes in the case of load imbalance.In the virtual machine livemigration process, it first need to select a node which has high load values and selectthe appropriate time to migrate it out. Selecting the "right time" is a kind of virtualmachine migration trigger strategies and appropriate trigger strategies can reducemigration times, save energy.Therefore, this paper combines cloud computing owncharacteristics, the paper design:(1) In the cloud computing environment, before the dynamic migration ofvirtual machine, in the process of resource monitoring, aiming at the lack ofreasonable charging function problems in the existing virtual resourcesmonitoring framework, this paper design acquisitionresource monitoring module, which can adapt to the virtual resource accounting incloud environment, generate a resource monitoring model in cloud computing couldbetter adapt to resources virtualization and charging characteristics.(2) In cloud computing environment, in a virtual machine Dynamic migrationprocess trigger strategy, to against the problem of existing single threshold policy andother simple dual-threshold trigger strategy instantaneous peak initiation migrate ortrigger timing unreasonable waste of resources, this paper presents the trigger strategybased on historical data using time series forecasting techniques and weighted dual-threshold, in order to achieve the purpose of reducing the number of migrationand saving resources.（3）Using Eucalyptus platform In CentOS to build virtual resourcemonitoring environment and using Linux command to get the usage of virtualmachine CPU, memory resource per second on a process level within300seconds。This fine-grained monitoring results resolve the issues that the service providerneed to deal with measuring the cost of using virtual resource by users in largeextent. This paper uses cloudsim cloud platform simulation software achievesimulation experiments, in the number of virtual machines and energy of virtualmachine migration both consumption,with a single threshold trigger migrationstrategies and dual-threshold strategy for comparison. Experimental results show thatthe migration strategy in this paper can effectively reduce the trigger virtual machinemigration times, reduce the energy consumption of virtual machine migration.