The Dynamic Migration Of Virtual Machines In The Cloud Computing Research

The fast development of computer software and hardware leads to the evolution of computing model.After distributed computing,parallel computing and grid computing,the computer industry and academe proposes a new computing model, cloud computing.By integrating networked computing resources as a virtual resource pool,cloud computing provides users massive computing resources as service.The service integration and on-demand supply provided by cloud computing can improve the utilization of computing resources and reduce per-service energy evidently. Moreover,it can shield the fault of computing resources.In cloud computing,resources are shared and can be reserved according to users'requests.Cloud computing platform has to guarantee to meet Service Level Agreement(SLA) if the users'requests have been approved.However,users' demand for resources usually changes with time.Eg,web server has different workload in different time.Reserve resources too early will inevitably cause low resource utilization.Furthermore,when users are sharing resource transparently,cloud computing platform has to do load balance to avoid hotspot.How to allocate resources efficiently and meet SLA at the same time is a big issue of cloud computing.System virtualization provides an effective way for resource management in cloud computing.By encapsulating services in virtual machines(VMs) which are finally mapped to physical servers,virtualization can do VM remapping according the workload.Live VM migration is a good way to do VM remapping.With the help of virtual machine monitor(VMM),which lies between operating system and hardware, live VM migration moves the whole execution environment,including operating system,from one physical server to another.It can help the administrator to do load balance,fault tolerance and online maintanence.Currently,live VM migration is initiated manully.And VMs can only be migrated between the same kinds of VMMs.In cloud computing,massive computing resources are integrated and may be managed by different kinds of VMMs.Moreover, while live VM migration is used to do load balance and meet SLA,the migration itself has to occupy system resources,including CPU and network bandwidth.The migration process may cause severe SLA violation without control.Generally, traditional live VM migration is slow,error prone and lack of control.Therefore, traditional live VM migration cannot acquire optimal resource utilization and meet SLA in the migration process.Moreover,traditional live VM migration can only be done between the same kind VMMs,which is lack of flexibility.Considering the features of cloud computing,we propose a solution for applying live VM migration to cloud computing.The solution mainly has three advantages. First,it monitors real-time resources utilization of the platform and initiates live VM migration automatically.Second,it can live migrate VMs between different kinds of VMMS.Finally,it dynamically controls the migration process to avoid the migration process possess too many system resources.To demonstrate the applicability of our approach,we have implemented a live VM migration framework,namely Vagrant based on Xen VMM and KVM.According to our experimental results,Vagrant can initiate live VM migration between KVM and Xen automatically according to system load.Meanwhile,it can control the migration process dynamically to meet SLA.Performance measurements show that the implementation is comparable to traditional live VM migration in end-to-end time, and the down time is less than 2 seconds in most situations.