A practical guide and methodology for Communications Service Providers on VNFs implemented as Virtual Machines.
Each VNF is like a snowflake. They are unique, and deployed and managed in completely unique ways.
Network Functions Virtualization (NFV) – running network functions in software, independent of any specific hardware – offers CSPs many potential benefits. These include speeding up service delivery, cost-efficiencies, and a simplified architecture based on open standards. As networks evolve to cope with massive workloads (such as IoT traffic) and new technologies (like 5G) the attractions of NFV are obvious.
But, as so often in the early stages of technology paradigm shifts, there is still a lot to be worked out before the process is optimized. Both CSPs and the Manufacturers who supply them have to balance innovation with pragmatism. As a result, the process by which Communication Service Providers (CSPs) and their suppliers shift more and more telco workloads towards the cloud (private or even public) is an evolutionary drift rather than a big-bang migration.
This evolutionary approach is also informed by the challenges of cost and complexity that have characterized the early experience of on-boarding VNFs to the CSP infrastructure. The foundation of delivering services faster, cheaper, and better is to leverage economies of scale, economies of scope, and faster and more complete learning curves.
HPE, Intel, Tech Mahindra, and VMware have, between them, on-boarded over 200 VNFs for Communication Service Providers.
CSPs are finding that on-boarding each of the dozens of VNFs they need to support their revenue-generating services is taking weeks or months, and incurring more than $100k in associated costs. One CSP executive observed that each VNF is like a snowflake. They are unique, and deployed and managed in completely unique ways. Aside from the non-standard nature and challenges in deploying VNFs, there is a myriad of potential implementation and automation technologies to choose from.
HPE, Intel, Tech Mahindra, and VMware have, between them, on-boarded over 200 VNFs for CSPs. Their collective experience has been consolidated into a paper, which seeks to offer learnings from their work. The intention is to assist CSPs in moving from NFV “automation” to NFV “at scale”.
This paper is intended to be used as practical guide for CSPs. It includes a methodology for on-boarding VNFs, the majority of which currently come in Virtual Machines. It includes a clarification of terminology, a phased on-boarding process, and details of the on-boarding methodology.
This methodology, focusing on VNFs implemented as virtual machines (VMs), will be current for most of 2017. We expect to iterate and improve this process by introducing a Common Data Model to describe VNFs in a uniform way.
An open-source orchestration solution based on published information and data models will unleash an OpenNFV ecosystem that addresses multiple use cases for carrier-grade network functions virtualization (NFV).