A Design Framework of Automatic Deployment for 5G Network Slicing

Differential services driven user-end and operatorend challenges have been the main driving forces behind the 5G network, which is well perceived as an innovative platform for digital convergence of information, control and management. With both the network slicing (NS) and service slicing (SS) technologies, precious physical resources can thus be shared among multi-tenant mobile virtual network operators, such as over-the-top (OTT) service providers. This paper proposes a three-stage design for automatic slice deployment called LMA, namely (1) LCP: local charm provision for VNF services, (2) MSP: model-based slice planning for service chaining, and (3) ASD: automatic slice deployment for flexible and virtual resource allocation. The LMA is a model-based slice-specific platformneutral design framework for deploying the NS and SS, not only automatically deployable on both the x86-based desk-top computers and data-center bare-metal servers, but also on public or private clouds, as well as the 5G mobile edge. Our design framework adopts the Juju-as-a-Service and EurecomMosaic5G software technologies, where several customizable virtual network function (VNF) components can be flexibly chained together to form a desired NS or SS. This paper studies and presents two successful deployment showcases: a web-blogdatabase based SS and a virtual-evolved-packet-core based NS. Our preliminary results of performance benchmarking show a strong effect of the number of CPU cores on the average latency response of SS, in particular during congestions caused by concurrent user requests.