Optimization Of VM Migration Strategy For NVM-DRAM Hybrid Memory In Edge Computing

With the popularity of wireless networks,massive heterogeneous terminal devices have continuously connected to the Internet,data types have become more complex and diverse,and the amount of data has increased dramatically.At the same time,smart homes.The emergence of a large number of new applications,such as intelligent networked vehicles,puts higher demands on the timeliness of services.As a new computing model,edge computing provides users with more real-time services by merging resources such as network,computing,and storage on the edge of the network near the user or data source.Because the user has mobility,to meet the low latency requirements of the service,the edge computing platform needs to migrate the virtual machine hosting the service to the edge server closer to the user.However,compared with traditional cloud computing,virtual machine migration technology in the edge computing environment faces the problem of node heterogeneity.Edge servers are increasingly composed of dynamic random access memory(DRAM)and permanent memory(PM).Mixed main memory,but does not take into account that virtual machine migration will generate many random write operations on the target node's memory,which will lead to memory performance degradation,and bring huge loss to the PM.How to efficiently migrate virtual machines to edge nodes close to users directly affects service delay,system computing power and resource overhead,which is a key research issue to guarantee service quality.In response to the above problems,the main work and innovations of this paper are as follows:First,the design and implementation of a virtual machine dynamic migration strategy for NVM-DRAM hybrid memory.This paper first re-examines the edge computing system architecture design and deeply analyzes the virtualization and dynamic migration technologies.Aiming at the edge computing platform with mixed memory,an optimized dynamic migration strategy is proposed.The LRU memory page transmission queue is designed to migrate the memory data,reducing the number of repeated copies of the same memory page during the iterative process of the traditional algorithm.The virtual machine mixed memory address is mapped to the physical memory address,and the hot and cold memory pages in the migration process are stored in the corresponding memory area of the destination virtual machine to reduce the loss of the PM and improve the memory performance.Second,the design and implementation of virtual machine hierarchical migration architecture under edge computing.The migration of virtual machine storage data is the bottleneck of virtual machine migration.In order to adapt to the WAN environment,this paper proposes a virtual machine hierarchical migration framework.By synchronously migrating the storage data on the source server to the target server,it can be deployed in advance.The data layer shortens the transmission time of disk data,reduces the occupation of network resources by the migration process,and reduces the migration overhead.