Balancing power consumption on the network edge
- By Stephanie Kanowitz
- Feb 04, 2020
Edge computing – bringing compute, storage and intelligence closer to where data is created – offers great promise for smart city initiatives, but the power consumption associated with it could butt up against municipal governments' efforts to go green.
The global IT power footprint for smart city infrastructure edge equipment was about 158MW in 2019 and is expected to reach 3,052MW by 2028, according to the “State of the Edge Report 2020.” More than half of that footprint is associated with smart building and public-safety solutions, and it’s expected to grow: Smart buildings are forecast to reach 36.9% of the overall smart city footprint by 2028, public-safety solutions will hit 24.8% and the remainder will be associated with traffic signals, smart lighting, public venues and utilities.
“We will continue to see a rise in power consumption as the edge spreads out” along with the rollout of 5G and the move toward software-defined infrastructures that telecommunications and wireless providers are embracing, said Scott Anderson, managing director of Packet Labs, a maker of small data centers that can be installed close to potential users. That proximity is important because it reduces latency.
“This is a way of decentralizing the internet, and it has an interesting tie-in with 5G and the type of rollout and the type of hardware and software that’s now being deployed as a result of cities moving into the 5G space,” Anderson said.
But the environmental outlook isn’t bleak. There are ways to tweak edge systems to minimize power consumption. For example, power use may increase when cities use draw on artificial intelligence to analyze data collected from telephone pole-mounted license plate readers and trigger them to home in on a specific vehicle. As the LPRs pass the data across a network to other compute devices, cities will likely spin up machines and instances, Anderson said, but the rest of the time, the power consumption remains steady because the city can run fewer instances.
Additionally, cities can set up that extra capability only in areas where problems persist. That way, the power needs stay relatively flat.
Another option is deploying software that would make compute resources available at specific locations, rather than citywide. For instance, Anderson said, managers of vehicle fleets may not need their application to work throughout the municipality. Having certain services and nearby compute power to compare data may be enough. That’s the service layer that could run right at the network edge, he added.
New York City is already taking into account this shift to software-controlling hardware., Mayor Bill de Blasio and CTO John Paul Farmer announced Jan. 7 the five-point Internet Master Plan. The goal is citywide broadband accessibility, but the plan also describes an “Edge Cloud,” which it defines as “a transition from a reliance on large data centers far away to having more, smaller data centers closer to the dense parts of a network…. An Edge Cloud can be as compact as a small room, with the proper environmental controls and power backup.”
The plan also calls “rooms prepared for network equipment … logical locations for small data centers, which are increasingly important and widespread as content providers put their content closer to their users and as network applications rely on edge computing power.”
Overall, cities can expect to see power consumption remain high, but management software introduces new opportunities to find out where the cheapest compute or smallest unit is, as opposed to spinning up a big server.
“It’s efficiency through software. That is the promise of what management software can do for us as we move into more decentralized places," Anderson said. "We can use smaller machines that have less power footprints.”
Jacob Smith, Packet’s chief marketing officer, was a co-chairman of the “State of the Edge Report 2020.”
Stephanie Kanowitz is a freelance writer based in northern Virginia.