Personal Services Placement and Low-Latency Migration in Edge Computing Environments

Low latency and integration with edge computing facilities are two of the main enablers of the upcoming fifth-generation mobile network. The geographical distribution of virtualized compute and storage resources (e.g., NFVI PoPs) allows to deploy application/service instances/components closer to mobile subscribers, and to provide much higher performance levels in terms of latency. These new degrees of freedom should be suitably managed in order to place and to migrate application components along with subscribers' move. Given the mass-scale of subscribers and the non-negligible overhead to move application components over resource-constrained edge facilities, this management should be performed in a highly scalable and effective way. In this respect, this paper presents the design of a comprehensive edge computing framework, including an orchestration algorithm, able to place and to move services in few milliseconds (<25 ms). The algorithm provides scalability levels able to support services instantiated on a per-user basis (i.e., personal services) also in the presence of complex network and service environments. The performance evaluation has been carried out by considering different service chains within diverse software live migration technologies.