By YuLing Chen & Alon Bernstein, Cisco Systems Inc.
In recent years, Network Function Virtualization (NFV) has been introduced into the Telecom industry to deliver reliable and efficient commercial networking services in programmable standard hardware systems, called Virtualized Network Functions (VNFs). NFV promises benefits in the savings of operational and capital expenditure (OpEx and CapEx), as well as the increased automation, operations simplification, business agility, and faster time to market.
The cloud native microservices container architecture was originated from the web scale providers such as Amazon, Google, and Netflix. The approach of cloud native is to break down a monolithic application into small microservices and deploy as containers in the cloud. One of the attractions of this approach is that applications can be tested in an iterative and distributed model, without taking applications offline. In the cloud world, large scale applications have been developed, tested, and deployed with more agility using this distributed model.
Since 2016, several large service providers have publicly embraced the move to a microservices architecture in the telco cloud. [1] There have been announcements from major service providers to use containers to build out their network function virtualization infrastructure. Some key telecommunications equipment suppliers are using microservices to implement some of their software. Open-source initiatives are moving towards microservices and containers. In NFV space, there is a trend of moving from the virtual appliance based solutions to the cloud native approach, which is referred to as the Cloud Native NFV.
The NFV world has been following ETSI NFV references. However, most of the ETSI published documents were based on case studies and Proof of Concepts built on virtual appliances. There is a gap between ETSI NFV and the cloud native approach. With more and more cloud native solutions appear in NFV, there is a need to augment the existing ETSI NFV specifications so as to continue guiding the NFV world towards interoperability and standardization.
To support this effort, this paper identifies the elements in the ETSI NFV Management and Orchestration (MANO) reference architecture that need to be adjusted when applying the cloud native approach in NFV. We also propose a pragmatic software architecture that realizes the NFV MANO functionality using the cloud native approach. With the focus on the network service design and deployment, which is the core functionality of the NFV Management and Orchestration systems, we exercise the TOSCA language for the service modeling in the cloud native environment.