Powering the new network edge

Edge computing and the new power generation

Power demands at mobile network cell sites and Central Offices could treble due to the demands from edge and distributed computing deployments.

Vertiv’s Alex Pope, Director of Product Management, Global Solutions, said that the power demands of Central Office sites could move from current typical levels of 3-50kW to 100kW while base station tower sites could shift from 2-3kw up to 10kW, and in some cases even reach up to 50kW.

That power demand will be driven by the installation of mini data centres at the network edge as operators install platforms to support their own  vRAN and distributed vCore deployments, as well as to provide co-location resources for other Cloud server, app and service providers.

In addition to meeting this new power demand, as operators evolve edge strategies their site infrastructure management and deployment will need to change.

Pope said that to meet these new demands companies such as his that provide DC and UPS power, thermal management, racks and enclosures are being tasked to match data centre infrastructure economics at the network edge, so that operators can efficiently upgrade and migrate thousands of sites.

That means producing full solutions including power, thermal, security all in a single package – that can be flexibly deployed to the Central Office or to base station tower sites.

“UPS and cooling demands will be high, but what has changed most fundamentally is the ability to make the transition fast, simple and flexible for these network operators at hundreds or thousands of locations,” Pope said.

“In the old model there was a lot more on-site work, a lot more one-off design per location, a lot of complexity. You might have different indoor/outdoor build materials, different installation teams working on site. Now the requirement is for packages that are very rapidly able to replace existing equipment at a smaller footprint but higher power density – for example a road-based self contained infrastructure including redundant power, thermal, and management security at rack level. These are solutions that we can standardise on, deploy at locations and then down the road add additional capacity as required.” 

“A lot of CPSs are looking at the right things that will work at a grander scale. What they want from us is flexibility so we can provide the base infrastructure that is efficient and reliable and can scale in a modular fashion. That way they have freedom to upgrade and add on at a late date.”

Pope did not say what the direct cost implications of this increased power demand would be, but he did say is that the requirement is to be able to match the economics of larger data centres.

“We can typically get design points of cost per kW that they are used to for larger datacentre applications. So yes there are increases from a 2kW shelter to a 30kW shelter but we are able to design in performance so that we can hit data centre based cost targets.

An archetypal approach

“That’s why we created the archetype classification,” Pope added, referencing research that Vertiv has commissioned that identifed the 24 most likely edge apps to have a significant impact. These were refined into four archetypes; data intensive, human-latency sensitive, machine to machine latency sensitive and life-critical.  Vertiv is building on this initial phase to define technology requirements for each archetype.

“We classified these archetypes because we believe the demand for those apps will drive the need for more IT assets at the edge – coming from different companies in different verticals. So if we can group those archetypes together and create patterns we can drive efficiencies of scale to have similar deployment, management and reliability requirements in the infrastructure.”

For example, Pope said, some categories would need very high levels of power reliability and physical security, while others would be less stringent.

“We believe that different requirements will surface regarding characteristics that relate directly to the infrastructure and others that may relate to the channel or go-to-market requirement. So we may have a distributed edge set of deployments supporting something life critical that will need to be extremely robust with multiple layers of redundancy and very high SLAs and onsite service requirements for repairs. Or we may have something data intensive but it if it goes down nobody dies. There we can be way more cost sensitive.

“In terms of go-to-market and ownership – you may have municipalities that want to own or have a level of control over life critical apps within their borders. That’s the same with financial or medical information in Europe – so there’s a stated customer preference for how assets are owned and controlled.”

Nor is this all something for the medium term future – some edge apps such as content caching and distribution are already placing increased demands on edge real estate, Pope said.

“It’s not just aspirational, we are doing it.” He also cited some health and smart grid applications, where edge processing is happening to determine if data needs to be sent back to a core site.

Vertiv’s classification of edge archetypes is interesting and provides a useful way to think about the drivers for edge deployments. However, the development of archetypal infrastructure solutions perhaps does not fully take into account the fact that mobile operators in particular may be delivering very different apps from within the same real estate space – and on common infrastructure. However there’s no doubt that as power demands increase, and site space is at an ever greater premium, more “productised” site designs that can match larger data centre economics will be of use to operator planning teams.