AI-driven network operations will reduce opex, streamline performance monitoring and increase operator agility
Fujitsu expects to see an acceleration in the pace of Open RAN (O-RAN) deployments this year and a consolidation in terms of commercial deployments in 2025, according to Patrick Eriksson, vice president and global head of radio unit business at Fujitsu.
“I think it [O-RAN] started quite well with a couple of greenfield operators that did fairly large deployments. After that, O-RAN deployments halted a little bit and then this step to come into brownfield deployments has been a little bit challenging to take. I think, from our perspective, what we’re seeing…right now is we believe things will happen this year, and come to fruition…maybe in 2025 and onwards,” he explained during a recent interview with RCR Wireless News at Mobile World Congress in Barcelona.
Given the parallel pushes for opex reduction and progress against net-zero goals, Eriksson also highlighted how Fujitsu is approaching energy efficiency (and attendant costs) at the hardware-level, and also based on an understanding of real world network demand.
“One is the raw hardware efficiency that we’re working with. And here we are utilizing, for instance, gallium nitride power amplifiers (PAs) throughout our whole portfolio. In addition to that, we’re also deploying proprietary PA architectures and energy efficiency algorithms that are helping to improve our overall power efficiency on the hardware side… we also have the features that optimize the power efficiency in accordance with the actual traffic load of the radio,” he said.
Using AI for network root-cause analysis and predictive maintenance
Fujitsu is also using artificial intelligence (AI) to develop new tools to help operators improve network performance. During Mobile World Congress, Greg Manganello, senior vice president at Fujitsu, explained the main features of the Alarm Storm Detector tool developed by Fujitsu using AI.
Alarm Storm Detector is an AI/ML-powered application engineered to streamline network performance monitoring. “We have combined our network operations knowledge with AI to really help network performance. And when there’s a network trouble, what we really realized is that it’s not a low level of trouble, but big spikes, which can be a little bit overwhelming to network operations teams. And we realized they’ve been spending days and weeks looking for the root cause. So using our AI tools, we can find the exact root cause of a network trouble, throwing away hundreds and thousands of derivative alarms and getting to the root cause,” said Manganello.
“The next step that we’re working on right now to release is predictive maintenance. So when we detect degradation of network parameters, we already present that to the operator,” he added.
Manganello also noted that Fujitsu is currently working on large language models—a sort of a chat GPT interface for network operators—, looking at log files and huge petabytes versus of data, and then helping the network operator query that database.
“But what we realize is LLM can’t operate on all that raw data. So we invented a new kind of AI called automatic knowledge extraction, which really is like a bridge between all that multi-vendor multi-domain data, and then network operators so they can get the answers back to their questions on how to really operate a network more efficiently,” he said.
