AI and network analytics are generating intense heat, and activity shows no sign of abating. In early July, Ciena purchased network performance and routing analytics solutions specialist Packet Design, an acquisition that is intended to support Ciena's ambition to provide closed-loop automation for programmable networks and to deliver what it calls the adaptive network. AI and network analytics also dominated this year's TM Forum Digital Transformation World, where a significant number of the case studies and catalyst projects addressed AI use cases, including automated or autonomous network and service operations.
CSPs have mixed feelings about autonomous networks given that it implies eventually handing over control of the network to machines, but that is not preventing the popularization of futuristic use cases. In this context it is important to remind ourselves of some of the underlying requirements that will need to be in place if autonomous networks are to become a reality.
The first of these requirements is network analytics capabilities that can support multi-domain network orchestration. With new networks and architectures adding ever greater complexity, there is an increased need for real-time insights on the performance at multiple network layers. Indeed, this need for multi-domain capabilities was one of the drivers of Ciena's Packet Design acquisition, and is the path being pursued by other forward-looking vendors such as NEC/Netcracker. Vendors that build out their network analytics capabilities to support multiple domains – such as RAN, optical, and IP – will find such solutions more attractive compared to those supporting single domains.
The adoption of 5G will further complicate matters, as it will require network analytics capabilities to also support sophisticated network slicing management and optimization approaches, handling a wide range of horizontal or vertical (customized) sliced offerings.
Standardized data formats will also be required. Before CSPs can start deploying autonomous use cases, standardized data formats and algorithms will need to be in place to handle the complexity associated with deploying AI technology within CSPs' existing siloed environments.
Also, to reassure operators, AI systems will need to be seen to function in an explainable way, providing engineers with insights as to how decisions are arrived at and the factors driving those decisions. Ideally, these systems should allow operators to validate the predictions and decisions taken. Operators will need to be able to analyze or replay events to ensure that decisions were aligned with organizational policies.
Having mentioned some of the obstacles that need to be overcome to realize autonomous networks, it is nevertheless worth stressing that quite a few CSPs are making strong progress with existing AI use cases. A good example is the work being done by SK Telecom to develop a next-generation OSS platform to leverage AI technologies for monitoring, optimization, and customer experience management use cases in its fixed and mobile network operations. Future development plans for SKT's TANGO platform include running it as an autonomous OSS platform, utilizing machine intelligence to provide zero-touch operations.
For more detail on SK Telecom and other network analytics implementations and use cases, see the following Ovum reports due to be published in August: Using AI to Improve CSP Network Operations: Use Cases and Using AI in CSP Network Operations: What CSPs Have Deployed.
Straight Talk is a weekly briefing from the desk of the Chief Research Officer. To receive this newsletter by email, please contact us.