Aria Networks, a pioneer in the field of AI and networking. It uses AI, machine learning and advanced modelling to automate and optimize networks.
The company specializes in optimizing multi-layer, multi-technology, multi-vendor networks at Tier1 CSPs, including BT, Level 3 and Verizon. At Mobile World Congress 2016, Aria featured in Vodafone’s proof-of-concept for virtualized service offerings, “VPN+”. While Aria is largely focussed on telecoms operators it has also recently done some work with Facebook on optimizing its global optical and IP network. This revealed a saving of 25 percent in its inventory of 100G IP ports.
In this Q&A session, CTO Dr Jay Perrett articulates his vision for the deployment of AI in telecoms networks and how AI can help build the business case for SDN and NFV.
TM Forum: Where are CSPs are today in terms how they are exploiting AI?
Aria: Many operators and vendors associate AI and machine learning with detecting patterns and predicting outcomes, especially in relation to customer data. But it’s only with the deployment of AI in the network – to make faster, more intelligent decisions – that operators have the opportunity to realize a genuine step change in speed and agility. An operator using AI the same way as a bank or a retailer – on customer transactions or behaviour – is not going to transform its business. If you’re not transforming how you operate the core asset in the business – the network – you’re not really transforming.
TM Forum: Can you tell us a little more about the benefits of AI in the network?
Aria: In transformation programs, the objective that is sometimes overlooked is making better business from your network. Today that connection between business and network relies heavily on humans to plan, analyse and decide how best to change the network. But as networks become more complex and dynamic, those processes won’t scale. AI is what allows operators to scale, closing the loop, by actually making decisions. And that’s important: AI is not just about finding patterns in data, it’s about taking action based on that information. That’s the key to growth – adapting intelligently to the network and business environment.
TM Forum: What are your thoughts about the current status of SDN and NFV deployment and where does AI sit in the migration to SDN and NFV?
SDN and NFV are about so much more than just replacing boxes. They’re not just another network technology. The real benefit is being able to operate and automate the new business models they enable. You can’t do that with SDN and NFV as operated in traditional networks. There has to be some kind of intelligence. The combination of SDN, NFV and AI make it possible – today – to have the network solve a business problem: How do I optimize the profit margin from my network? How do I reconfigure the network to restore services in the event of a DDoS attack? What’s my worst case failure scenario and how do I mitigate it? I want to describe what I want, not specify how to get there. The AI bit figures out how to get there. AI is essential for SDN and NFV migration.
TM Forum: Where do you see AI playing a role in 5G?
Intelligent automation is crucial in a 5G network. 5G is about enabling a new model for telco, in which devices rather than humans are the focus. The network will be able to instantiate new services for networked devices at a moment’s notice. Those services may be short lived – a few minutes or hours. Imagine a maintenance drone backhauling video to a data centre. That is going to mean automated decision-making on the network: when, where, and how best to create or modify network slices. The network needs to take into account traffic load and current failure states at the time as well as mitigating for predicted changes in both. There’s a lot of interest in using AI to make the 5G network smart enough to decide the best option for itself.
As part of a TM Forum catalyst we did a 5G slicing demonstration with Ericsson, Viavi and WeDo, sponsored by AT&T and NTT. That was a solid example of figuring out how you get from service catalogue to slices being created and managed in real time without human intervention.
TM Forum: How important are standards now that we are seeing various reference architectures coming into play?
Aria: There are various reference architectures around, from AT&T ECOMP to Vodafone’s Ocean, with a big focus on orchestration. However, MANO does not answer the question of where and how decisions are made. That’s why operators are having to go beyond MANO to build something that connects with the business. That creates the need for a new standard for integrating analytics and optimization intelligence, and we’re working with TM Forum and Vodafone to drive that forward.
Thinking about standardisation and AI more broadly, as soon as you fix a design [as part of a standardisation process]you have baked in what you can do with the components of a network. You need to make the design more flexible to benefit from AI. We should move away from having people look at detailed designs and let that be done by a component with intelligence. The high level design and the building blocks are what we should be working on and these are the components that AI should be able to connect up.
TM Forum: So, bringing these various strands together, what would be your advice to telcos on AI, SDN and NFV and next-generation OSS?
Aria: Ultimately, it’s all about the business. The goal is not virtualization – it’s automation. How well can you automate the process of adapting the network to satisfy demand, while turning a profit? AI is the only way to arrive at answers fast enough, at scale. So it’s important not to limit your thinking on AI to dashboards and customer data: it’s critical to how CSPs will scale next-gen operations in a software-defined future.