Telco AI solution for energy efficiency delivers 20% reduction in power consumption
A recurring theme in telco AI discussions is around acknowledging the grandiose transformational potential of the technology while embracing the practical reality that AI success is likely to start with narrow, practical use cases built on top of a well-managed data platform. From there, steadily layer in new use cases and ideally these small, incremental advancements ultimately set a flywheel in motion that powers an AI-enabled, highly-automated machine.
But, as is the case with many types of new technologies, accelerating adoption often hinges on how easy the technology is for the user to consume in service of business-specific goals. “We’ve been in AI/ML and talking with customers about it for four or five years now,” Blake Hlavaty, global director of network software offers at Fujitsu, explained. The conversations have evolved from vendors delivering a set of tools and leaving the user to put them together to providing technology that’s packaged in service of delivering an outcome. From the CSP perspective, “There’s all this great technology but I’m busy operating my network. How do I bridge the gap from great technology that has the potential to help me versus use cases?”
With its Virtuora portfolio, Fujitsu is aligning with the move toward open, cloud-native networks by providing interoperable telco AI solutions for high-value use cases, including alarm storm detection and root cause analysis, anomaly detection, energy efficiency, network snapshots, traffic prediction and more. The next step is to add in Virtuora Intelligent Applications which couple neural network modeling with pre-trained models built using domain-specific data. These applications are also compatible with third-party LLMs.
While 5G monetization efforts slowly come together, Hlavaty said he sees CSPs focusing on using telco AI for network optimization and opex reduction. “The use cases on network optimization are more clear than what’s going to help an operator monetize the network.” AI investments are being focused on areas “where you can show clear, definitive value.”
To give an example, there’s an industry focus on using AI to optimize RAN energy efficiency thereby decreasing power costs. Fujitsu has an rApp that runs on its Virtuora SMO; it uses AI and ML to estimate network traffic then switch network capacity on or off as needed while maintaining service continuity. Testing shows a 20% power savings. Discussing how to drive adoption of this solution, Hlavaty looked at it as indicative of the larger issue of developing trust in AI systems in the march toward network automation.
“We’ve got to rewire how we think about operating networks if we’re going to be able to keep up,” he said. “New people are going to come in that don’t have the history that others do and look at things differently. Our approach to that is…we’re not going to start with closed loop. Let’s just help the person make decisions. Let’s start there and build that trust.” As they gain familiarity, CSPs will gain trust and take the step from open loop to closed loop automation.
“We’re probably a long way from that full automation closed loop on any kind of large scale or for any major problem,” he said. But that’s still the direction of travel. Hlavatay reiterated the need for an incremental approach to identifying and solving particular use cases. For Fujitsu, that means using its domain expertise “to bridge the gap between the technology being developed on the AI side and the needs of the network operators. We’ve got to get creative and we’ve got to try stuff.”
