Research On Deployment Strategy And Mechanism Of Virtual Machine In IaaS Cloud

Cloud computing can provide users an efficient, self-service, elasticcomputing resources and application service. VM technology has been appliedto cloud computing and can provide more optimal utilization of resources inexisting infrastructure. Cloud computing providers can provide the flexibilityservices for more individual users. VMware, Xen and other products haveapplied in cloud computing widely. The VM technology has become animportant technology of cloud computing.The VM technology is already widely used in cloud computing. In thistechnology, physical conpute and storage resources can be integrated into a largeresources pool of cloud computing platform. The pool is divided into manyindependent VMs, and these VMs provide users services as traditional physicalmachine did. These virtual machines are isolated from each other, and user caninstall different operating systems, applications and services in these. In order torespond quickly to changes of user loads, the resources which virtual machinesand clusters can be dynamically expanded or reduced.An important feature of the cloud computing is flexible. When the user’sloads change, the platform will be able to respond quickly by expanding orreducing computing and storage resources required dynamically. The way to thefeature is a rapid deployment and controlling the size of the VMs. The VMsdeployment includes selection of the host, images transmission and VMscontextualization. These issues have a great impact on the cloud platform performance and user experience. How to deploy virtual machines efficientlyand reduce migration time is one of the focuses of the current virtual machineapplications in the cloud.This article focuses on the dynamic host selection, virtual machine imagetemplates transmission and contextualization. At present, some works associatedwith this study has begun, but still need to studied and optimized deeply. Maincontributions in this article are as follow:(1) Study on the mechanism of the host selection. An individualized hostselection mechanism was proposed. It predictes multiple loads of the host in theselected range dynamically, then weights the resource requirements of the virtualmachines which will be deployed, finally dynamically selects and matchs theappropriate destination host according the prediction value of the load.(2) Study on the mechanism of the virtual machine images templatetransmission. There are some defects of performance and redundancy in existingtransmission mechanisms. A multi-point transmission method based onredundant storage technologies was proposed. Firstly it makes a pool by anindependent division of storage space on each host. In deployment, each nodecan employ multi-point parallel transmission. Thus the pool can disperse trafficflow and reduce single point load and can make template data redundant forensuring the reliability of deployment.(3) Study on automatical contextualization of virtual machine. A method isproposed for contextualizing automatically virtual machine. A contextualizationproxy script is putted into the virtual machine image template. After transferingthe templates to selected host, in the instance process, first the virtual machinesvia DHCP or stateless address configuration protocol to obtain a dynamic IPaddress to access network, and then access to contextualization informationfrom the contextualization service by the agent. To ensure the safety of thecontextualization information, the asymmetric key encryption algorithm is used. (4) The above studied contents are applied to an application called Web6to4.This application provides users with Web6to4service. When the number ofusers is increasing, it is rapidly deploying virtual application servers in the cloudand spreading the load with more and more users’ access. The test results showthe deployment method of the study can effectively shorten the deployment timeof the virtual machine.