Virtualisation is only the first step to NFV success

Mobile operators in search of a ‘software only’ NFV solution are likely to be disappointed, as hardware still has to be considered, warns Robin Kent, director of European operations at Adax

Virtualisation is only the first step to NFV success

Network Function Virtualisation (NFV) has continued to grow and prove popular across many industries in today’s connected world. Telecommunications is one such industry that is reaping the benefits it offers.

According to a whitepaper from the Linux Foundation, the main goal of NFV is to increase service agility while enabling better asset utilisation’. However, this goal often gets lost in the search for a ‘software only’, NFV/SDN solution without really understanding the needs of individual networks and applications.

To prove successful, NFV can only occur when there is architecture in place for both hardware and software. Considering software in isolation doesn’t work. Individual networks will vary according to local demands and operator preferences, and an appreciation of the roles of both hardware and software is essential for the NFV architecture. This is especially true when considering applications that require a simplified hardware assist solution.

Hardware acceleration
The performance demands of certain applications and network functions, such as security gateway, Deep Packet Inspection (DPI) and the 4G/LTE core network, require hardware acceleration on the data plane because of the high data throughput of these applications.

Separating the control and data planes improves the control of the network and paves the way for a more effective and reliable NFV implementation. But it’s not enough to stop there.

This separation will centralise the control plane so that third-party applications can run autonomously, which turns the control plane into a Virtual Network Function (VNF), leading to a more flexible and dynamic network. Yes, agility and flexibility are improved, but also the OPEX cost to the operator is reduced by simplifying the control and programming of the network.

The best practice for operators should be to move the control plane to a virtual machine, offering a simplified solution for network management while adding to the quality and value of the overall solution with hardware acceleration for the data plane.

Resource requirements
Operators need to ensure they have virtualisation and distribution management system resources in place. The role of a Virtual Network Function Manager (VNFM) is a necessity to establish close communications between the NFV manager, and therefore the software and hardware in the field.

Too much of the focus for operators is on searching for a virtualised network solution that is on generic COTS hardware, without understanding the actual network or user requirements. In fact, hardware and software can and should work in tandem, but it’s important to recognise how and where that hardware runs in the network. For now, applications that require hardware assist require dedicated and specialised COTS hardware.

With a control plane that’s both centralised and virtualised, the next question is how to ensure that the applications can meet the demands of performance, reliability, flexibility and scalability when faced with high throughput.

Performance demands
As a result, the performance demands of certain applications, network functions and nodes such as a Security Gateway, DPI and the 4G LTE core network, point towards the need for hardware acceleration at the data plane.

Today, it’s important to take advantage of both COTS and dedicated hardware, and make sure the control and data planes are decoupled for NFV. Not all applications need hardware acceleration but many do, especially those with high packet throughput.

Separating the control and data planes simplifies network control and operators need to have the hardware solution in place to accelerate and deliver cost-effective, reliable application performance.


Leave a Comment