Saving energy, in a network context, ultimately means that somewhere along the line, user experience could be impacted. Perhaps coverage or throughput ticks downward, or users are shifted from one cell to another.
In a session at the recent Telco Sustainability Forum virtual event (available on-demand here), Jen Hawes-Hewitt, head of strategic programs and solutions for the global telecom industry at Google Cloud, and Yannick Martel, VP and data and AI lead for the telecommunications industry at Capgemini, discussed the balance that operators need to strike between competing interests of reducing their energy usage when user experience expectations are high.
“It’s not a simple case of, how do we simply reduce energy spend,” Hawes-Hewitt said. “There’s a real, delicate balance that has to be struck constantly with the quality of customer experience and the ability to stay competitive. Because one of the real challenges that our CSP customers see is, if they compromise on customer experience and have network drops, call drops, have lower throughputs, et cetera, then they are going to have churn,” she continued, adding that network operators need to “retain just enough power in the network so you’re able to have a fantastic customer experience, but do that in the most effective way” using real-time data feeds. Hawes-Hewitt said that, given that the Radio Access Network consumes roughly 70%+ of network energy costs, the RAN is the focus as the first place to apply cutting-edge techniques driven by artificial intelligence and machine learning, in order to save energy.
That means saving on OpEx cost as well. Energy costs are “not just a big part of the spend, it’s also an increasingly volatile part of the spend,” Hawes-Hewitt said. Operators don’t just want to reduce overall costs, they also want to reduce the volatility in their energy bills. One of the major challenges, however, is dealing with the existing, data (and power) hungry legacy tech stacks, and data siloes between different technologies, vendors and domains that prevent holistic configuration changes that could better optimize networks.
Generally, when the industry talks about achieving energy savings, they are often referring to doing so in 5G and future 6G systems, said Martel of Capgemini. “But indeed, we want to achieve savings now, with the networks we have today,” he added. CTOs first turn to network equipment manufacturers to help with optimization. “We feel that really, you can go beyond [that],” he said. To do so, “you need to get data out of the network and to get data into an open environment where you can really understand the way the network is working, the way customers are using it, and how the consumption is distributed—and then, you can really find levels to activate and to change configuration, maybe, and to save electricity.”
The first phase of identifying such opportunities is by using historical network data for simulation purposes, which can help identify and prove out changes that could be made for energy savings without actually touching the network. Updated data can come from the network itself. Telecom networks are incredibly complex, and there are many ways they could be mis-configured. Understanding patterns of activity and usage is crucial, Martel pointed out, and simulations can enable operators to identify changes and savings and build a business case for those changes. “You really need the simulation to build a business case for the project,” Martel continued. “One the business case is sold, you know how much can save, you understand where you want to play. And then you can decide whether the savings is enough and decide for the investment.” The idea is that eventually, operators could eventually progress to live data feeds from the network to make optimizations in a more real-time manner.
Bringing network data into simulations to demonstrate energy savings opens up the possibility of exploring that data for other uses, Hawes-Hewitt pointed out. “When you see the operational savings you can make with RAN energy savings, you also look at maybe service assurance, or some of those additional important use cases that are also reliant on increasingly more AI/ML, more automation,” she added. “At this stage, we’re still seeing a fair amount of human interaction to validate the configurations, to validate what [are] the appropriate fixes to put into the network. But that’s also a really interesting progression over time to an increasing autonomous network.”
