Sunday, April 7, 2019

Lean NFV proposes a Key-Value Store to simplify deployments

A new industry initiative known as Lean NFV is aiming to simplify the process of onboarding virtualized network functions (VNFs).

Six years after the publication of the first Network Functions Virtualization (NFV) specification, proponents of Lean NFV argue that the technology has failed to live up to its promises largely due to complexity and inefficiency in the deployment model.

The topic was addressed in a keynote at last week's Open Networking Summit in San Jose presented by Scott Shenker, Professor of Computer Science, UC Berkeley; Sylvia Ratnasamy, Associate Professor of Computer Science at UC Berkeley & CTO of Nefeli Networks; and Constantine Polychronopoulos, VP & CTO of the Telco Business Unit at  VMware. 

Specifically, complexity in the current deployment model arises in three areas:  when the NFV manager is integrated with the existing computational infrastructure, when VNFs are integrated with the NFV manager, and when coordination is required between the various components of the NFV manager.

Lean NFV is advocating a key-value (KV) store to serve as a universal point of integration for VNFs. Many open source key-value store implementations have been deployed by companies such as Google, IBM, LinkedIn, Redis, etc.

The goal with Lean NFV is to allow NFV solutions to be “plugged in” to any computational infrastructure. With this model, principles, an NFV manager is responsible for the management of both individual VNFs and end-to-end service chains, including lifecycle management tasks such as placement, launching, configuring, chaining, scaling, healing, monitoring, and upgrades. KV-based integration also allows VNFs to evolve incrementally and independently.

Lean NFV argues this model is synergistic to ETSI MANO and supportive of the work being done with ONAP, the network automation platform of the Linux Foundation. Furthermore, the model scales for the network slicing and microservices requirements in full 5G architecture.

A whitepaper on this topic is published here: