While the industry attention to NFV has focused on virtualising machines, many of the benefits of virtualisation will be missed if CSPs don’t deploy predictive analytics to allow proactive attention to network management and service delivery based on customer preferences
As network functions virtualisation (NFV) starts to roll-out much of communications service providers’ (CSPs) attention has been focused on the benefits of deploying commoditised hardware and a little attention has been given to NFV MANO (management and orchestration). However, amidst the excitement, the telecoms industry has yet to devote sufficient attention to the intelligence that is required to feed orchestration and effectively manage virtual network functions.
“NFV is an emerging technology and most of the efforts have been expended on virtualising the functions, the mechanics,” confirms Suren Nathan, the chief technology officer of Razorsight. “The what to do and when – the intelligence – are what is coming next.”
It’s as if CSPs are putting the cart in front of the horse by deploying the cart first in the form of virtualised equipment, then considering the horse and reigns in form of MANO and service control, when actually they need the intelligence of the cart driver to decide what functions need to be instantiated.
“You have to put the intelligence up front and for that you need analytics but we don’t see that being considered yet,” adds Nathan, who advocates that CSPs adopt a cloud-based predictive analytics platform to optimally manage customer preference inputs and automated data feeds from the rapidly expanding plethora of IoT (Internet of Things) devices to maximise the potential operational efficiencies and economic benefits of NFV.
Nathan explains that there are series of compelling business cases for predictive analytics in support of NFV. Among these is predictive capacity planning and forecasting. In contrast to traditional capacity planning where a mobile operator would, for example, take market analysis from the marketing department to predict future growth and data from traffic studies from the network department to augment that and then decide that additional capacity would be required in six or 12 months, predictive analytics provides greater accuracy and more flexibility.
“The traditional approach means infrastructure has to be planned and budgeted for months in advance so you don’t get flexibility,” says Nathan. “Once you invest in capacity, you’re done but, with predictive analytics you can predict in advance what the demand is going to be at a given point. CSPs can use the high degree of accuracy of those predictions to turn on capacity at the time they require it.”
“That has a big impact if you are able to delay investments based on actual need rather than hypothetical need,” he adds. “When virtualisation is linked to that, the real power of NFV becomes apparent.”
A second business use case is in preventative network maintenance. “If you think about reactive failover, the time when the failure happens to turning on another machine can be substantial,” says Nathan. “Predictive analytics can tell you precisely when a failure is most likely to happen with a high degree of accuracy – for example a weather change, event surge or a new technology introduction, you can proactively use NFV to change over without having to wait for failover. We’re talking about fixing problems proactively before they occur.”
Chris Checco, the president and chief analytics officer of Razorsight, adds: “If you can predict in advance when specific network elements are going to fail, you can get down to very discrete use cases.”
A third business case entails delivering an enhanced customer experience. “If you look at a football stadium during a big match, or traffic congestion resultant from an accident or inclement weather, you can predict there will be lots of upload and download traffic and ensure network capacity is instantaneously made available to support the burst of activity rather than reacting to it once it happens,” Checco explains.
Finally, Nathan puts forward the business case of service level agreement (SLA) management. “One of the benefits of NFV is the virtualisation of the enterprise network and CSPs are looking to include that as part of their offering and include aggressive SLAs,” he says. “The ability to maintain and guarantee a high degree of service availability backed by a strong SLA is a business differentiator. Delivering this depends on the ability to turn things on and off if anything goes wrong. Predictive analytical insights are critical here because any risk of an SLA impact needs to be proactively managed.”
Nathan is adamant that a cloud-based platform is the way for CSPs to go. “Cloud-based predictive analytics fit hand-in-glove with the dynamic nature of NFV and the connected universe driven by IoT and increasing volumes of mobility in software,” he says. “When we talk about predictive data science and analytics these are new technologies. The NFV model is cloud so cloud-based predictive analytics makes perfect sense.”
“In addition, thanks to the cloud model, the analytics works on all the data – the CRM, the network data, IoT connected devices and the customer interactions,” Nathan adds. “A traditional approach to network engineering would only consider a limited data set while ensuring capacity and availability, which never factored in important customer related aspects including loyalty, churn propensity, demographics, social media implications and myriad other factors. Analytics correlates the information to address the business need.”
However, this level of analytical insight can’t be achieved by data scientists that are not telecoms experts. “Telecoms is at a crossroads,” says Nathan. “While the network changes, the market place is changing and customer expectations also change. This requires a thorough understanding of what is happening so in this dynamic environment, it is going to be very difficult for someone with no understanding of the industry to generate critical analytics.”
“Industry experts who are also data scientists and software engineers like the team at Razorsight will drive much greater value in coupling NFV with advanced analytics than generalists would,” he adds.
Nathan sees the situation becoming only more complex and therefore regards intelligence and insights as vital for CSPs looking to maximise the benefits of virtualisation. “Analytics is a critical component that must be bundled into NFV roll-out,” he confirms. “If CSPs want to be agile, they have to be proactive and utilise the power of real-time, predictive analytics.”
“We are waiting on the underlying orchestration tools and APIs to mature,” he concludes. “It’s an evolving area but the more the industry couples NFV and analytics and understands the remarkable potential on differentiation through vastly improved, proactive customer experiences the better the market will be.”
Adds Checco: “One only has to look at how Netflix has redefined the video consumption market via proactive analytics as an interesting parallel.”
