The present and future of network automation
As 5G continues its evolution toward broad digital enablement of the enterprise, the proliferation of artificial intelligence and machine learning in networks will be a crucial part of service provider strategies. For infrastructure, increasingly automated deployment, configuration and management accelerates time-to-revenue as closed technology stacks give way to open, cloud-native architectures.
Here we’ve compiled commentary from carriers and vendors around the role of AI- and ML-based network automation today, and what the future holds for network autonomy.
Morgan Stern, vice president of automation strategy with automation software specialist Itential, framed it like this in an op-ed published by RCR Wireless News: “CSPs are facing more pressure than ever before for their networks to perform without failure. With the explosive growth in cloud-application usage, mobile devices, and extending connectivity and computing to the edge with 5G, CSPs must keep up with demands for more total capacity and higher throughput. This shift forces network operators to change how they manage and deliver network services.
“As a result, network operators want to employ automation to help respond to increased demand with faster provisioning, while maintaining service delivery requirements. Since most CSP networks span multiple domains (cloud, mobile, wide area network, radio access network, edge, transport, data center, managed services, etc.) and require a higher level of investment and management, ensuring network automation initiatives are correctly designed, developed and implemented is key to future success.”
Rakuten Mobile Chief Technology Officer Tareq Amin said, “I can see an absolute trajectory to the analogy of what [the]car industry is doing today to achieve full autonomous driving. I see it.” He’s leading his organization toward an “autonomous network in which my operation team doesn’t have to do a thing to run and manage this network. I actually tell my team, ‘I want you to be lazy, my friends. I want you to be lazy by making sure that what you do every day is about code. Network-as-a-code is what your mind needs to be programmed about. Elevate your intelligence. Push the envelope.’”
Amin said the network needs to change, the operations support system needs to change, and the mindset needs to change. He said the goal is to build “a very simple marketplace that allows the [user]to consume both IT and network applications…We want to completely transform all of this.”
“It’s not like AI and ML are foreign to the operators today,” Microsoft’s Shawn Hakl, partner for Azure Networking and a veteran of Verizon’s enterprise group, explained.
He pointed to current uses, similar to the SON capabilities outlined above, such as “automated recovery and automated resilience…and then automated policy settings. They use AI and ML to predict network failures and analyze network data.” But, and this is a significant but, “The thing is,” Hakl said, “it’s not connected to taking any action.” He characterized some processes as operating “semi-autonomously” but, by and large, painted a picture of a reactive approach to network operations. The next step is to start leveraging that data to make decisions in closer to real time. He saw security-related processes as a likely candidate for more and more automation. “People are already working on advanced threat detection, automating sandboxing of bad behavior. What I see is people just narrowing down the time between detection and reaction then automating that up to do it at scale.”
Currently, adding intelligence to the radio access network is the focal point of AI and ML adoption by telecom operators and their vendors. And given the increases in traffic and complexity brought about by the advent of 5G, alongside the move from proprietary to virtualized/cloud-native network functions, this makes perfect sense.
According to Dell’Oro Group Vice President Stefan Pongratz, “The increased complexity with the various 5G technologies in combination with the shift towards Open RAN will potentially introduce new challenges to CSP operational teams tasked with managing end-to-end performance. Artificial intelligence will play an increasingly important role managing this complexity deliver the quality of experience (QoE) that consumers and enterprises demand from mobile broadband applications and latency-sensitive services.”
Tantra Analyst Founder and Principal Prakash Sangam laid out the parallel developments of distributed 5G networks and artificial intelligence, a world where AI processes that previously took place in a centralized cloud are now being distributed throughout the network and on to devices. He looked to a future of fully-distributed AI.
“There has been a clear trend towards distributed AI,” Sangam explained. “It’s now cloud as realized on device…for very good reasons. A lot of use cases are there where you need immediacy. Also there is a move toward edge cloud. My view is the AI will be distributed between cloud…as well as edge cloud. Most of the data that AI needs to work on is generated on devices. It would be not feasible and not smart enough to send all of that raw data to the cloud for processing. That’s the reason why you need distributed AI crunching that data at different levels based on the need of the use case, of the application, and kind of data.”