NFV allows CSPs to increase efficiency and lower costs, but it requires a transformation in the way networks are run, writes Kate O’Flaherty
A virtualised network is useless on its own: it needs the structure provided by orchestration. But NFV (network functions virtualisation) orchestration requires a radically different approach to traditional CSP network management. This is seeing network and IT departments converge in order to meet evolving data and analytics needs.
In the NFV environment, network equipment is no longer dedicated hardware and instead comprises servers and switches. The technology therefore needs to use the capabilities developed in cloud computing, such as automated provisioning of compute and storage, automation of network provisioning and service chaining in the data centre.
It is not a simple transition, but the benefits of NFV orchestration are worth the complexity. Like cloud, NFV is easy to scale, removing the risk of service unavailability due to capacity shortage. Therefore CSPs do not have to provision large, spare capacities; this can instead be allocated on demand.
Orchestration frees up capacity, which can then be used for other services, by mapping the instantiation of virtual network functions against real-time demand, says Joachim Mason, head of data centre, UK and Ireland at Cisco Systems: “CSPs can also achieve a more predictable and optimised return on investment by deploying additional network services without unnecessary equipment costs.”
When used in the right way, virtual tools can be configured to perform any network function, says Ronnie Neil, customer experience assurance product marketing manager at JDSU. “We can quickly introduce capacity when required or launch a new service. This adds efficiency, as well as lowering cost and increasing agility in terms of doing things faster.”
According to Shaul Rozen, director of product strategy at Amdocs, orchestration is “the brain of the new network”. He says: “Think about what would happen if you just virtualised the network and did not orchestrate it: the end point would be just the network, but virtualised – which is no good.”
Orchestration challenge
However, orchestration faces multiple challenges. Services are not currently joined up and sit across multiple vendors and standards. Rozen says: “One of the key challenges is creating an ecosystem that can be applied and certified on top of orchestration. Another is the current lack of standards: different vendors offer varying templates depending on their own functions and capabilities.”
There is also a lack of definition of what a virtual network manager should do, says Rozen. On top of this, no decision has been made on how many orchestrators are needed and whether there should be a hierarchy such as global and local variants of the function.
Appropriate data and analytics is essential for making configuration decisions. According to Rozen, there are three types of analytics needed for smart orchestration: network analytics to deal with congestion and predict traffic flows; compute power to assess availability of resources around data centres; and customer analytics to use information for better business decisions when orchestrating.
For example, Rozen says: “If you need to prioritise one of two services, one way to go about it is to look at customers that are high value or pay more.”
Without the right tools, operators do not have transparency and control over a distributed NFV platform, says Andreas Lemke, the senior marketing manager for cloud at Alcatel-Lucent. “With NFV analytics, service providers can detect potential issues before they become critical. For example, by observing resource utilisation trends, an NFV platform can raise alarms and take proactive measures to avoid bottlenecks.”
IT transformation
The technology is proving increasingly beneficial, but in order to take advantage, CSPs must transform the way they are structured. Whereas within traditional CSPs, operations runs the network and IT handles the infrastructure, a virtualised environment requires the two departments to work together.
IT departments are usually the pioneers of using virtualisation, so are in an ideal position to assist, Rozen points out. “There is a need to shift skills: IT capabilities need to be transferred into network teams – and there should be a leakage of personnel between those two departments.”
Peter Christy, research director at 451 Research‘s networking practice agrees, saying: “Successful NFV solutions are a blend of IT and network competences. Neither can succeed without the other.”
This, combined with other issues, means NFV deployments will take time. Due to a lack of standards and the state of the technology, it will still be at least five years before NFV is in action.
In the meantime, many CSPs are looking at a phased approach to deployments, says Neil. “They don’t expect to have whole networks virtualised. It will come in stages with the simplest parts first. For example, let’s implement virtualisation for a single network element.”
As the ‘brain’ of the network, there is no doubt that orchestration is integral to NFV’s success. However, the scale of transformation should not be underestimated. As Neil points out: “If you compare the introduction of virtualisation to 2G, going to 3G, to 4G, it is an order of magnitude bigger. It’s a dramatic change.”
