Font Size: a A A

High-performance Container Live Migration System For Mobile Edge Computing

Posted on:2022-10-29Degree:MasterType:Thesis
Country:ChinaCandidate:Z W ZhaoFull Text:PDF
GTID:2518306572990959Subject:Computer system architecture
Abstract/Summary:PDF Full Text Request
In the mobile edge computing(Mobile Edge Computing,MEC)environment,the contradiction between the limited coverage of a single edge server and user mobility may cause the user's service performance to decline or even service interruption.Using container live migration technology to migrate users' services to the edge server near the user is an important way to solve this problem,but the existing container live migration solution is difficult to ensure fast service handoff in the MEC environment.Aiming at two key issues that affect the performance of container live migration in the MEC environment: migration destination decision and migration time overhead,a highperformance container live migration system is designed.The system adopts the following schemes to complete the selection of the migration destination node and reduce the migration time overhead: decentralized migration decision model,memory snapshot scheme,adaptive pre-copy strategy and spatio-temporal parallel data transmission scheme.Among them,the decentralized migration decision model selects the appropriate migration destination node based on the network delay between the mobile user and the destination edge server and the free resources of the destination edge server,which reduces the communication time overhead in the process of selecting the migration destination node under the condition of ensuring accuracy.The memory snapshot scheme uses memory to store snapshot data,avoids additional hard disk I/O overhead,and provides support for data parallel processing solutions.The adaptive pre-copy strategy reduces the amount of snapshot data to be processed during the migration process by reducing unnecessary snapshot operations,and avoids additional system overhead.The spatio-temporal parallel data transmission scheme uses data parallel processing,pipeline processing,and adaptive compression strategies to speed up the data transmission during the container live migration process and reduce the downtime and completion time of the migration.Therefore,the highperformance container live migration system can solve the problems caused by user movement in the MEC environment and ensure fast service handoff.The test results show that,compared with the centralized decision model,the decentralized decision model reduces the average response time of user services by about7.2%,and reduces the communication time overhead in the migration decision process by about 52%.Compared with the latest container live migration solution proposed for the MEC environment,this system reduces the downtime of container live migration by10.1%?63% and the completion time by 16.7%?51.8% under different network bandwidths.
Keywords/Search Tags:Mobile edge computing, user mobility, container, live migration
PDF Full Text Request
Related items