Mark Bramfitt - Utilities & IT

Greenpeace Evaluates Cloud Providers on "Cleanliness"

Greenpeace has published an updated report that attempts to evaluate the environmental "cred" of cloud service providers, and has certainly generated lots of press coverage (there was a fifteen second bit on the local news here in the Bay Area last night).

The report attempts to estimate the power footprint for the data centers of fifteen cloud service companies, and coupled with an evaluation of the power content provided by local utilities and whether they publicize and share energy efficiency practices, comes up with a letter grade for energy efficiency and greenhouse gas mitigation. Letter grades are also on offer for "energy transparency", infrastructure siting, and renewables and advocacy.

Both Data Center Dynamics and the New York Times have issued stories on the report, including reactions from some of the leading companies that were evaluated. DCD commentary runs to the defense of the operators; the NYT is straight up reporting so far.

I think I've got shots all around: for the report and for the cloud service companies, and even for DCD.

Let's start with the report. Greenpeace acknowledges that in quite a few cases it had to estimate the data center loads for the cloud providers, and in the case of Apple, they seem to be in a tit-for-tat about whether the estimates are reasonable. (Greenpeace says that Apple's data center in Maiden NC has a load of 100 MW; Apple says it is more like 20 MW but won't say exactly. I suspect Apple may be at 20 MW now, but is rapidly fitting out the DC to its eventual full-load capacity that has been widely assumed to be 100 MW.)

Instead of the arguments, I would urge the companies to divulge their data center loads or annual energy use, which in turn would earn them higher marks in the "energy transparency" category.

My serious quibble with the report is that it rates coal-fired and nuclear generation in the so-called "dirty category", which is indefensible in my view. You may not like nuclear for other reasons, but it doesn't emit greenhouse gases, so shouldn't be part of the evaluation. And natural-gas fired generation, while producing half the GHG emissions of coal, should certainly be a included. 

As to the cloud providers, it's past time to "come clean" about what they're up to. Searching for cheap power rates means you're going to get fossil-generated power - you make the decision, live with it. And while I applaud the (very) modest moves by some of the companies to explore on-site renewable generation like photovoltaics and biogas-fuelled fuel cells, we can all agree that these are rather more in the category of demonstration projects rather than meaningful mitigation programs.

Look, eBay and Google come out well in the report, in part because they acknowledge the impacts of their operations, and primarily because they're willing to pay more than lip service to real initiatives to reduce those impacts. 

As to Data Center Dynamics, or for that matter any data center industry group or analyst, state your criticisms of the report and your sugestions of how to analyse these issues more fairly. You have to give Greenpeace all due credit as one of the only organizations that is trying to do that.

Apples and Oranges (Demand and Energy) Strike Again for Solar at DCs

As always, interesting news from Data Center Dynamics revealing the extent of solar photovoltaic on-site generation by eBay, including a new installation at a data center in Utah.

First, you simply have to commend eBay for the scale of their commitment to on-site generation: the new installation brings them close to 1.5 MW of solar PV, with another 0.5 MW in fuel cell capacity. To which I say "hear hear".

But the reporting does get a bit sideways when it comes to the numbers. eBay states that their "renewable energy generation capacity is...11% of the company's total data center power demand" (emphases mine).

That's exactly how I'd write it if I wanted to extoll the virtues of the projects, but I'll join Amazon's James Hamilton, who took Apple to task for their solar generation claims, in spinning this quite a bit differently.

While fuel cells can produce power around the clock, we all know that PV arrays have a much more limited operating cycle. So the correct comparison, especially against a block load like data centers, would be on energy production, not demand.  I would expect that the amount of energy produced by eBay's solar and fuel cell plants equates to something like five or six percent of the energy used by their data centers.

I haven't been able to find out if the Bloom Energy fuel cells at eBays Silicon Valley campus are using natural gas from the local utility or landfill gas (they are near a landfill). If the former, the power can't be called renewable, and if the latter I wouldn't call it renewable either. (The gas from the landfill will decline unless you continue to expand the facility, and let's be real, calling it renewable would be a corruption of the term.)

Lastly, I find it very hard to believe that the solar plant in Utah will pay for itself in only four years, given prevailing power rates there. It's on the order of seven or more years in California (after accounting for subsidies) where power rates are far higher.

Again, all due credit to eBay for their investment and commitment - it truly warrants attention and commendation. But have your engineers check the attempts by publicists to spin the data: it diminishes your accomplishments.

 

 

DCD on Apple's North Carolina Data Center

Bit of news reported by Data Center Dynamics on Apple's Maiden, North Carolina data center - I take particular interest as I am doing some consulting work with the local electric utility, Duke Power.

The squib is all about Apple's plan to install almost five megawatts of fuel cell capacity at the center (mistakenly referred to in the opening paragraph of DCD's piece as "cogeneration", unless Apple plans to do something with the low grade heat from the plant).

It appears that Apple plans to sell the plant output to Duke, rather than to use it to offset the data center load. This isn't necessarily a surprise.

If you have a self-generation facility, it can be more lucrative to sign a power purchase deal with a utility rather than offsetting your own use when you carefully assess the economics. For example, the utility will likely charge you a standby service fee if you use the power in-house to cover the infrastructure needed to serve you should the generation plant go off-line.

There might also be power purchase subsidies in play, which make the arrangement more lucrative.

And finally, there is real value in choosing when to run your plant. It would appear that running around the clock would amortize your investment the quickest, but off-peak power rates may actually be cheaper than your cost to produce power, so running only in on- and partial-peak hours may be better.

All by way of saying that I'm sure Apple has analysed the economics that support the purported power sale arrangement.

Speaking At: ConEd Energy Efficiency Summit 2012

Well, not speaking so much as moderating a panel discussion...

ConEd has a great program planned for their 2012 Energy Efficiency Forum on May 1 in midtown, and I'll be on hand to moderate an afternoon session on deploying data center energy efficiency easures to capture capacity.

We have a crowded panel, with both vendors and end-users, so will likely fill up our forty-five minutes, but I will encourage our panelists to be available after the session to answer questions.

Really looking forward to supporting this event, and if you have facilities in New York City, you should think about attending!

Via DCD: Emerson Lists Ten Data Center Trends

Great story in the DatacenterDynamics newsletter today (a must subscription) relating ten data center trends noted by Emerson - the power delivery and conditioning folks.

Ambrose McNevin of DCD claims not to be surprised by the list, though I have to say a few raised my eyebrows - at least in terms of raising more questions.

Take the first one: rack densities will increase to an average of 12 kW within three years. I certainly agree that if so, data center operators are going to be challenged with power delivery and cooling issues. It would suggest that airflow management will be a necessity, if not a move to completely isolated airflow, which will represent a big investment in legacy facilites.

The second one doesn't ring true, or certainly hasn't in the recent past. Data center operators replace power delivery and air conditioning systems with every third IT refresh cycle? Every ten years? My experience is that cooling systems have a fifteen year useful life. And I've seen UPS systems that are half my age!

I'll save one last critique for item ten, which states that IT workloads vary a lot by time of day, week, month, and even year. The examples certainly ring true, but with most users loading IT gear at ten percent or lower of capacity, these workload swings are essentially meaningless from a power or cooling perspective. Look at the utility meters, as I have, and most data centers have the closest thing to a base loading profile as you can get, with any variation more likely attributable to weather conditions (affecting cooling efficiency) than any other factor.

AESP Paper and Slide Deck: Desktop Computing EE Opportunities

Thanks to those of you who attended the panel session at the Association of Energy Service Professionals Forum in San Diego where I delivered a presentation regarding desktop computing energy efficiency opportunities. (I appreciated the feedback too!)

The presentation covered what Alan Rose from Intel and I describe in a paper - what the energy efficiency opportunities are, and how utility programs should be designed to drive market adoption of the technologies and measures.

I know you'll be able to get the paper and (eventually) the presentation deck from AESP, but I'll post them here as well:

Opportunities for Desktop Computer Energy Efficiency Programs 

Mark Bramfitt, P.E., Bramfitt Consulting, San Francisco, CA and Alan Rose, Intel Corporation, Portland, OR

Download Bramfitt AESP Paper 12 12 11

Presentation Deck:

Download AESP Pres Deck Bramfitt Rose

Speaking At: Electric League of PacNorthWest Data Center EE Summit

I'm looking forward to a full-day training session in Bellevue WA on April 11th, where the Electric League of the Pacific Northwest is hosting a Data Center Energy Efficiency Summit.

The bulk of the day will be spent describing energy efficiency opportunities in data centers, though as always I'll provide an industry review at the beginning, and close the day with a call to action for the data center operators who attend.

Thanks also to four local utilities in the area, who are marketing the program to their customers. We also expect to make time for three or four customer case studies during the day, which always brings the training material to life.

I've always been impressed with the commitment to energy efficiency in the Pacific Northwest, evidenced by truly superior utility programs, and am excited to be a part of a renewed focus on the data center market there.

There's Still a Market for Utility-Scale DCs in High Power Cost Regions

Two reports out this week show that there is a market for large data centers even in areas with high power rates.

The first covers the new Vantage data center in Santa Clara - a 9 MW wholesale colo featuring outside air economizers and a reported PUE of 1.12. It's the second of three data centers that Vantage is building, and can be scaled up to an 18 MW load.

The GigaOm article also reports that Vantage has their own substation, so is able to purchase power on the wholesale market, which presumably means they can best Silicon Valley Power's retail rate for large customers of about eight cents per kWh. 

That's not really competitive though when some utility-scale data centers are getting power for under four cents a kWh in some markets. 

But it does show that there is still a big market for data center space in Silicon Valley despite the costs, and I'd point out that Vantage's center was built in just eight months - incredibly fast. Kudos to Silicon Valley Power for being able to serve a load that quickly.

The other report from the New York Times describes the conversion of a former telco high rise in Manhattan into a data center - something we've watched companies like Verizon and AT&T do in-house.

The project developers claim to have 40 MW of power availability at the site, and no knock against Consolidated Edison, but they'll be getting perhaps the most expensive juice in the country.

There's a good discussion of other data center properties in New York in the article, along with a description on how rental rates are structured. As expected, rent isn't based on floor area but power capability: $300 or so per kW of capacity per month, and then you pay for the power.

My quick back-of the envelope calculations tell me that to house one typical server in a colo in Manhattan would cost about $3000 per year. We used to say we'd passed the point where the power costs for a server over it's useful life exceeded it's purchase price; in this case you surpass it in just twelve months!

Government Giveaways To DC Developers - Why On Earth?

Datacenter Dynamics reported last week that two states - Nebraska and Iowa - are evidently in a shoot out to attract a $1.2B data center development project, and at least Nebraska is falling all over itself to offer attractive tax treatment. (I make that a 60 to 80 MW project.)

Maybe "attractive" isn't the word - Nebraska is willing to extend an existing program for large developments to data centers, absolving them of property taxes and providing refunds for local property taxes, and exemptions from sales taxes on equipment purchases and state use taxes.  To which I ask, what taxes does Nebraska plan to levy on the project?

They're even considering legislation allowing public power districts to sell so-called "excess power" to large projects at lower rates. I wonder what that means...

As always, the stated rationale is to attract high-paying tech sector jobs, but if you've been to utility-scale data centers in the past, you know that there is plenty of room in the parking lot. There just aren't that many employees, certainly after the site is up and running and close to fully utilized. I've been to 40 MW data centers that have no more than twenty employees on site.

Industry colleagues have noted that it isn't their fault that cities, counties, and states bend over backwards to attract major data center developments, but that is as much a cop out as saying the industry shouldn't be held responsible for the carbon content of the power it buys. 

If a company wants to build a data center in a community, they should pay their fair share of taxes to support that community and to mitigate any impacts of the project. Sure, seek out low power prices, but build to the highest efficiency standards and stand ready to implement carbon mitigation strategies. 

Just ask Facebook how it feels to be held accountable for these standards by their customers. That treatment will be extended to everyone building these large data centers eventually, so it would seem prudent to me to carefully consider whether loopholes are your best strategy today.

 

SuperNAP is as Big as Vegas Gets

I often refer to "utility scale" data centers in presentations and education forums because the scale of these facilites drives home the magnitude of the impact they are having on utilities and energy use in this country. (I do give credit to the operators of utility scale data centers for driving the industry emphasis on energy efficiency.)

Until now, I've called out data centers that just exceed 100 MW of load as the behemoths in the segment, or the data center "parks" that have loads of 250 MW or so.

Now comes word of an expansion of the Switch SuperNAP facility in Las Vegas in a report in DatacenterDynamics newsletter: when complete, the facility will exceed two million square feet, with room for 31,000 equipment racks. And power? Right at 500 MW, with equivalent generator backup (that's 250 units, assuming 2 MW per).

Perhaps my standard way of converting load figures into layman's terms are now out of date. Energy use by SuperNAP could serve 375,000 homes in California, but maybe we need to start comparing these power draws to full-fledged cities, or oil refineries, or total loads of familiar utilities...