Why business objective driven optimisation powered by AI has the lead over big data analytics
Posted on: 14th April 2016
By Wolrad Claudy, CCO at Aria Networks
Many of the conversations around the impact of virtualisation on network efficiency have been based around using data from the past to help shape insight into future demand. This process means collecting and analysing data to look for previous trends that might shape ongoing capacity requirements. But this kind of Big Data analytics has issues. It is a complex process, demanding huge amounts of time and computational power. Big data analytics builds an OSS/BSS overlay to the service delivery platforms, adding increased complexity. Many organisations suffer long delays from the start to the end – rendering the insight gleaned at best to be out of date and at worst fairly useless.
Furthermore using data from the past will not necessarily help clarify issues in the future. Service providers that are launching a new data service need to know the impact on the network of the service, but historical data patterns will not help with this. In other words, big data analytics does not provide the answers, just data with which to make a decision.
Generating profitable telecommunications
Telecommunications networks are, at their simplest, a delivery medium for digital content. People want to watch Netflix and surf the internet and the telecommunications network lets them do this. The reality today is that almost all revenues are based on the premise that you receive an almost unlimited volume of data for a fixed cost. One customer might have an ‘all you can eat’ tariff of which they use 500MB a month and another (paying the same cost) might use 10GB.
Similarly the cost of a megabyte of data may vary from more than $500 per MB (in the case of SMS at 10c per 160 bytes) to less than $0.005 per MB (assuming a customer is subscribing to a mobile broadband tariff at 5GB of data per month at $25).
With so much uncertainty, it is hardly surprising that telecommunications service providers struggle to understand which customers are profitable and which network investments will yield the best returns.
Business driven optimization
Ultimately what is required is an optimization decision engine that is able to understand what the network needs to achieve – both now and in the future – and deliver what we at Aria Networks describe as Business Driven Optimization (BDO). BDO is a dynamic and predictive value engineering function in the control plane of the network.
BDO aims to put control back in the hands of the CFO by enabling a clearer overview of network performance, not simply against technical considerations but also against business constraints, such as margin, depreciation and return on investment. Ideally Aria Networks believes that, in addition to business and technical considerations, the CEO/CFO should be able to optimise the network against regulatory constraints too, such as carbon footprint or percentage of population covered by service.
Enter artificial intelligence (AI)
Most scenarios will include multiple considerations – such as margin, service level and power consumption. Combining these criteria creates an incredibly complex but equally powerful network – one that is optimized in an optimal way to respond to a very specific set of criteria.
Artificial intelligence (AI) as a machine learning process is capable of very quickly creating two different scenarios from which a SW application can judge the best option against a number of constraints, such as the business, technical and regulatory criteria that matter to service providers. Once the best scenario has been chosen this will be compared to another and another until a best possible solution is achieved against the criteria set for what’s defined to be ‘optimal’.
In short AI is extremely good at self-learning to solve an incredibly complex problem quickly.
AI – the only route to release the potential of SDN and NFV
This is why AI is so fundamental to delivering the self-optimizing network of the future. Only AI is fast and flexible enough to be able to sift through billions of different network scenarios to quickly find an optimal network configuration against multiple business, technical and regulatory constraints.
The introduction of virtualisation can have an impact on network efficiency against different criteria, but only AI can deliver the transformative effect needed to optimize a network that can react to different scenarios in near real time.