We know now that Network Functions Virtualization (NFV) has been the biggest change to hit networking in the last decade. This year’s NFV World Congress, held in the Silicon Valley, offered an opportunity to examine developments from market leaders, debate the issues with thought leaders, and see the reality of NFV first-hand.
Network operators are restructuring their procurement and operational strategies to take full advantage of this fundamental change in underlying technology. Service providers are now moving into the “doing” phase of NFV, supported by a strong momentum of ecosystem partnerships, public Proof-of-Concepts and – for the first time – open source communities such as the OpenDaylight and OPNFV projects.
During the first day of the NFV conference, I participated in a panel debate on reliability in the NFV world. We debated a number of questions, including whether 100% service and application availability can be achieved, how does the new NFV paradigm affect the traditional view of reliability, and whether there is an opportunity for innovation.
As I argued during the debate, I believe there is a change in the way network operators are addressing reliability. Unlike the traditional paradigm where network operators simply ‘drop’ more boxes into the problem to increase service reliability with redundancy, in the NFV cloud-based world we can instantly instantiate multiple Virtual Network Functions (VNF) as needed to provide higher service reliability.
In the NFV world, my colleague Shai Herzog likes to think of VNF resources as plain paper cups. What do you do with a leaking or torn paper cup? You throw it away and simply replace it!
In the NFV world, service applications are designed to tolerate the failure of virtual machines, lifecycle is measured in hours to months and the SLA requires the ability to create/destroy virtual machines instantly as needed. The known equation of ‘higher availability equals higher redundancy’ can no longer be applied in the NFV world. Though in theory we could instantiate multiple VNF for ‘hot standby’ redundancy, in reality as NFV cloud-based solutions are real time environments, it is simply unnecessary.
This change in the way network operators preserve reliability is enabled primarily through constantly reduced prices of cloud based computing, storage and networking resources, versus today’s proprietary equipment and software that are needed for new value-added services.
I believe that in the next few years we will see more VNFs that were designed for a dynamic cloud based network (see – Pets versus cattle discussion). The reliability of these services will be based on near real time continuous fulfillment (which is founded on customer orders, real-time status reports and KPIs), that continuously evaluates the difference between the VNF service’s desired state and the existing state. With any change, it automatically will recalculate the difference in real time and re-fulfill the service to achieve the desired state.
This new concept for continuous fulfillment enables higher reliability and service availability in the NFV world, allowing for service providers to treat VNFs as plain paper cups that can be very easily tossed and replaced, as needed.
As Amdocs Head of Network Cloud Service Orchestration, Dr. Eyal Felstaine has spearheaded the establishment of the NFV Unit. An innovator and entrepreneur by nature, Eyal has co-founded and held CXO positions in several successful start ups in the communications and Evolved Packet Core space. An experienced speaker at industry conferences, Eyal has authored numerous articles, referred journal papers, patents and internet standardization documents (RFC).