UPS As A Service Could Be An Economic Breakthrough

Published on 28th February 2014 by Mark Monroe

Uninterruptible Power as a Service

How much would you pay for uninterruptible power?

Wait, let me rephrase that… how much are you paying now for uninterruptible power, and how much would you be willing to pay to have someone else manage that?

Data centers and uninterruptible power go together like peanut butter and jelly, and it would almost be heresy to propose building a data center without provision for UPS, generators, switchgear, and all the operational mess that goes along with providing electricity to computers on a continuous basis. Facebook’s Jay Park is known to have tried to build the Luleå, Sweden facility without generators since the Swedish grid had not failed once in 35 years, and feeds from no less than 3 independent generating stations were available on the site. Yet even bleeding edge Facebook could not bring itself to build a data center without generation and ride through capability.

So, each data center builds its own infrastructure for uninterruptible power, purchasing all the accoutrements needed to be their own power station for the short periods in the data center’s life when the venerable grid fails to fulfill its duty of supplying quality, continuous power to the energy hungry IT equipment inside the facility.

Cost of Power Conditioning

Power conditioning infrastructure is one of the most expensive and longest lead time items in the data center construction world. According to Schneider Electric’s Total Cost of Ownership tool[i], electrical infrastructure can range from 11 to 45 percent of the total construction cost of a data center. In a 1MW facility with a construction cost of $10M, that’s $1.2M to $4.8M, averaging about $2.6M / MW.

Over a 10 year life of the electrical equipment, total cost of building and operating that electrical infrastructure will be between $3.5M and $6.5M. Let’s use $5M total cost for easy math. Assuming a median PUE (1.5), 7x24 operation, and $0.10 per kWh electric cost, this facility will consume somewhere around 130M kWh over 10 years, at costs of over $13M .

Allocating capital and operating costs over the lifetime means the uninterruptible infrastructure is costing between $0.0269 and $0.0500 per kWh. The average cost we selected above comes out to $0.0384 per kWh.

Considering that most data centers are paying commercial or industrial rates for electricity of $0.066 to $0.098 per kWh, uninterruptible power for a data center adds a 27 to 77 percent premium to the cost of the electricity delivered to the IT infrastructure.

Virtualization Applied to Power

Now suppose that a reliable party, with sufficient assets or insurance to back up any claims against them, offered you uninterruptible power for a premium on top of the regular utility price. In other words, you sign an agreement with the party that guarantees whatever level of power reliability you want, and you pay for it as you go, rather than building the capability yourself.

This is, in essence, what co-location companies do all the time. They build and operate the infrastructure to provide power at the reliability level contracted in their Service Level Agreements, and the customers who occupy the racks inside the facility take the power provided, sometimes receiving compensation if the co-location provider doesn’t meet the terms of the SLA.

What if the co-location providers bought uninterruptible power for $0.01 to $0.02 per kWh from the utility, or a third party who builds common power conditioning infrastructure, manages the systems professionally, has economies of scale, innovative design, higher efficiency and utilization, and provides the same guarantee that the co-location provider would if they built the infrastructure themselves?

A facility that provided uninterruptable power to a cluster of data centers could also provide services to the power utility, like voltage and frequency sag protection, renewable or conventional power firming, ride through for gas peaker startup, any number of services that could improve the economics of the facility while providing the uninterruptible service.

In markets with variable electric pricing, it’s possible a centralized UPSaaS facility could become a short term generator, hedging the data center customer’s peak pricing, peak loads, and carbon content by producing electricity with natural gas. In states with high amounts of coal-based electric generation, a natural gas generator close to the consumers could be a big advantage from an economic, environmental, and energy security standpoint.

Advantages of a Service

Like most of the “as a Service” monikers in the industry today, there should be some serious advantages to centralized uninterruptible power as a service. For an area with 10 data centers in relative close proximity (Ashburn VA, for example, or London), in a traditional UPS build, each data center would build N+1 or N+2 or even 2N capacity inside their facility, consuming floor space that could otherwise be rentable.

In addition, the extra capacity of the “+1s” or “+2s” sits idle and unused, an insurance policy against untimely failure of one of the required N devices. And we’re talking the full infrastructure here: extra generators, switch gear, UPSs, battery strings, everything. This means in aggregate, we are really 11N+13, since each site has a “+1” just in case.

In a UPS as a Service site, the redundant needs can be combined and constructed more efficiently. The generator field can be 11N+2 rather than 11N+13, saving the capital and OPEX of 11 sets of power infrastructure while still providing the required level of service.

There could be additional service offered that was not economically possible before; if a portion of a facility has a requirement for 2N redundancy, the service provider can flexibly provide the required uptime, adding feeders for new levels of reliability instead of additional equipment.

Having a group of professionals focused on meeting the SLA at the lowest cost will also provide an advantage. A centralized facility will have better purchasing agreements for the aggregated service, since they will buy much more equipment from the supplier than the individual data centers. Maintenance contracts, fuel deliveries, security, personnel requirements, all will benefit from the economies of scale.

This could be a good business for the unregulated side of a utility, in states that allow that. The UPSaaS provider could meter energy usage by each client, pass through billing information to the utility, even offering collection services where it makes sense and is allowed.

Aggregation Helps Utilization

There are obviously a lot of considerations to work out before the concept of UPSaaS is realized. One reviewer of the idea expressed concern over being able to get air quality permits for a large collection of generators in one location. The counter to that concern is that the centralized service actually has fewer generators than the aggregate individual sites, so air pollution should be lower in total. Fuel storage, water use, security, noise pollution, all these and other issues might be easier to deal with at a single site than at a group of individual sites.

The big question is, would any data center owner hand over the risk of reliable power to a UPSaaS provider? Well, having your own plant is no guarantee that you will be able to provide reliable power service. The chart below from Emerson’s 2011 study[ii] on the cost of outages shows roughly 30% of data center outages are caused by failures of the owners’ electrical backup infrastructure, the same infrastructure those owners would be reluctant to give up because they don’t trust others to run a reliable service.

The question is similar to the one asked of IT executives as cloud infrastructure became available, reliable, and economically compelling. The combined impact of lower construction cost, additional revenue space, and higher efficiency should be enough to compel owners to consider the idea. Not all companies are able to take advantage of leading edge ideas. But the ones that see opportunity and seize it are frequently rewarded by riding the profit wave ahead of the others.

References

[i] Schneider Electric. “InfraStruxure™ Total Cost of Ownership,” http://www.apc.com/tools/isx/tco/

[ii] Ponemon Institute. "Calculating the Cost of Data Center Outages." Ponemon Institute sponsored by Emerson Network Power, 01 Feb 2011. Web. http://goo.gl/QkS0hf .

This blog was co-authored by Jack Pouchet, Vice President Business Development and Energy Initiatives, Hyperscale Solutions, Emerson Network Power. Jack works closely with large data center users, leading mission critical engineering firms, and utilities to help define, architect, and create opportunities for advanced energy, power, and cooling technologies and systems that improve day-to-day operational efficiencies within the data center eco-system.

 

CONNECT WITH US

Sign in


Forgotten Password?

Create MyDCD account

Regions

region LATAM y España North America Europe Em Português Middle East Africa Asia Pacific

Blogger

 

Whitepapers View All