Mark Bramfitt - Utilities & IT

EPA Utility Data Center EE Program Guide

I'm surprised I haven't posted on this before, but I was proud to be a part of a team that has developed a Utility Data Center Energy Efficiency Program Guide that has been published by the Environmental Protection Agency's Energy Star Program.

The guide truly distills the best practices for utilities that want to promote energy efficiency for their data center customers, delivering cost-effective results. The report doesn't paper over the challenges in this market (getting trade allies on board, addressing split incentives, free-ridership risks, etc.), and does suggest numerous ways to overcome these market barriers. I like to point out that running a program for this market is no more difficult than those addressing industrial market sectors, and often just as effective.

I have also been participating in monthly webinars for utilities that want to get a taste of what the report covers, and we've had great response rates (almost one hundred participants for the first on last year, and thirty on the call this morning). 

As soon as I can find the EPA's site for signing up for the webinars, I'll post it here.

Designing a Utility Incentive Program for Server Virtualization

It appears from market surveys that server virtualization has become a mainstream technology: most operators of enterprise data centers have tried it and are moving to full adoption pretty quickly. 

At the same time, in the two-and-a-half million or so server rooms and closets in the US, there isn’t much data available, though in a small study and anecdotally server virtualization and consolidation doesn’t appear to have much traction.

In California, where the major investor-owned utilities offered incentive programs for virtualization, the Energy Division, a unit of the California Public Utilities Commission devoted to protecting ratepayer interests, has ordered the end of these programs. (They were extended for some public sector customers, but have now completely ended.)

That decision has cast a pall over the entire country, with utilities deciding not to debut programs even though they might have poor market penetration in their service areas.

And more to the point, that atmosphere is leading to reluctance by utilities to address the server room and closet market, where a utility program could make a real impact. 

Just as the decision to extend programs in California for public sector customers was in part a recognition that market penetration was lagging there, is there an opportunity to offer a program specifically targeting server rooms?

There are a few of ways to design such a program: limit customer participation by market segment, or by rate class (which can be a proxy for customer size), or perhaps simply by project size.

While the first two would work and would capture small and medium sized business, for example, they would exclude corporations that often have a large concentration of server rooms “hidden” in their real estate portfolio.

Offering a program that paid an incentive for the first ten or twenty servers removed per site as part of a server virtualization consolidation project would draw in the broadest swath of potential customers. If delivered through IT service providers, the program could truly target the “hard to reach” (that’s a standard utility program term, by the way) server closet market.

And though this program could still reside under a data center and IT market effort, there is some benefit to consider folding it into a commercial office program portfolio. Server rooms can often be a significant portion of the energy use in commercial office facilities but have never been addressed through targeted market programs.

It’s a stretch to think that commercial programs, which address lighting and space conditioning equipment, could encompass server room technologies and measures as well. But if you did, and included desktop infrastructure and office equipment as well, you’d truly have a comprehensive program offering.

The upshot is that while the IT industry focus on data center energy efficiency has been vigorous for the past six years or so, server rooms and closets are only recently gaining attention. Utility program managers have an opportunity to affect this hard-to-reach, dispersed infrastructure using innovative program designs that could legitimately include incentives for server virtualization.

Virtualization Market Penetration

The market for virtualization software is proving something of a conundrum for utility energy efficiency program managers, leading to the cancellation of incentive programs in California, and hesitation by utilities in other states to debut program offerings.

What’s the problem? Utilities are bound by program evaluation principles that prevent providing incentives for technologies that have reached market saturation, or in other words for technologies or measures that customers are implementing on their own without incentives.

So answering the question about market penetration becomes crucial for utilities and program evaluators, and in the case of virtualization, the market statistics paint a murky picture.

I’ve sourced two rather different pictures of the market. The first is a “virtualization industry quarterly survey” prepared by Vanson-Bourne, a market research company, and posted at V-Index.com.

Their latest data indicates that in the US in Q3 2011, over 43 percent of all servers are now virtual, and that almost ninety percent of the companies sampled expected to increase their use of virtualization in the future.

The 500 plus companies that Vanson-Bourne sampled are all “large-scale enterprises”, though significant portions of that sample are located outside of the US. 

(Another interesting set of data from Vanson-Bourne indicates that most end users expected a ten-to-one or so consolidation ratio, but actually achieved on average more like five-to-one. A ratio that low would certainly yield far lower energy savings per project, and the move to newer, more efficient servers would become more of a driver of the savings.)

My second set of information comes from an informal survey done by the Natural Resources Defense Council under the leadership of Pierre Delforge, who is looking at the opportunity for energy efficiency gains in server rooms and closets.

These “distributed data centers” represent that vast majority of facilities, with an estimate of two and a half million compared to twenty thousand or so enterprise data centers. Market research indicates that almost sixty percent of the servers in use today are in these small facilities.

NRDC’s sample size was quite small (about ninety companies and institutions), and they don’t hold it up as a rigorous study, but their results are very interesting.

Of the organizations surveyed, only 37% had virtualized at least one server, and more starkly, less than a quarter expected to pursue additional virtualization projects. 

Even combined, these two data sets paint an incomplete picture with regards to adoption rates of virtualization. You could certainly conclude that most big users have implemented some level of virtualization and are rapidly moving towards saturation levels. 

You could also conclude that small and medium businesses have a much lower adoption rate, and are moving more slowly to make virtualization a standard practice.

What you can’t truly discern is the “squishy middle”: are large, sophisticated users adopting virtualization in their enterprise data centers, but not in their distributed IT infrastructure (server rooms and closets)?

So where does that leave utility programs? In the next post, we’ll look at California, and a design proposal for a utility program to promote virtualization in the distributed data center segment.

 

 

 

SVLG DC Summit: Server Room/Closet Efficiency

As you might expect, I thought that the panel session I moderated at the SVLG DC EE Summit was the best, though I will concede it wasn't the best attended!

Our topic was driving energy efficiency in the server room and closet segment, and we started off with a presentation from Bill Tschudi from Lawrence Berkeley National Laboratory, who shared what they found after auditing a dozen or so server rooms in a diverse customer group.

Bill noted that there were good low-cost opportunities for improvement at every site they surveyed and that the penetration of virtualization/consolidation was very low. LBNL has issued a three-page guide noting fourteen energy efficiency technologies and measures for server rooms (I was on the team that wrote it.)

Next was Shalini Singh, a project manager at Stanford University, who described how they tackled energy use at two of their over fifty server rooms.  In one case, the six or so racks of equipment will move from a server room with a PUE of over two to one of their data centers where the PUE is in the 1.4 range, saving over $100,000 a year in power costs.

In the second server room, supporting facility operations, the room will stay in place but a virtualization/consolidation program is under way that should address existing cooling issues.

Last but not least was Pierre Delforge from the Natural Resources Defense Council, who gave us a look at the opportunity nationally (we could save the equivalent of 250 MW of generation resources if we capture cost-effective measures), and touted our work at proposing a utility program design to address this market.

Participation from the attendees was good, with a strong acknowledgement that the opportunity is very real, but a challenge to address. Pierre and I are still working with utilities who are interested in running a program pilot, and we're also looking for IT service providers who would be interested in supporting such an effort.

SVLG DC Summit: ABB Reviving Direct Current?

Another solid Silicon Valley Leadership Group Data Center Efficiency Summit yesterday at AMD in Sunnyvale - thanks to Anne Smart of SVLG for inviting me to attend and moderate a panel discussion.

More on that in the next post, but first some notes from an interesting session on bringing direct current power distribution back to the data center market.

It turns out that ABB has purchased a major stake in Validus, the start up that had the strongest play for the technology, and that HP is in the mix by offering plug-and-play DC power supplies for a pretty wide swath of their data center IT equipment offerings.

Lutz Boettger of ABB noted that direct current technology was in some ways being held back by the lack of switchgear for DC, something they are bringing to the table in the new partnership. They have an expectation that market traction will take place in telco centers (which need to move away from 32 volts DC to higher voltages due to capex costs), modular data centers, and of course greenfield construction (this is not a retrofit technology!).

With a solid case study from a data center in Switzerland that offers a side-by-side comparison to standard AC distribution designs, ABB noted about a ten percent energy efficiency gain - rather more modest than some earlier claims out in the market. But they also noted a 15% capital cost savings, and a 25% footprint decrease, and these are certainly attractive benefits.

HP is bringing out a power supply that "slots in" to many of their server models, as well as to some data storage and networking gear. They're debuting another DC power supply that will work for blade chassis next year.

Bloom Energy was part of the panel also, confirming that their data center play rests on obviating the need for UPS gear and back up generators, and that DC distribution supports that strategy. (Fuel cells produce DC.)

From a utility program perspective, supporting data center new construction projects that feature DC power delivery and conditioning schemes will be difficult, for the same reason that supporting chiller-less designs using free cooling is: the project will have a lower capex cost compared to traditional designs, so justifying a financial incentive becomes impossible.

That's a pity, because the market will likely need quite a push to ge going in the US.I expect some sophisticated players who operate utility-scale data centers may be early adopters, but moving this down into the enterprise DC space will prove more challenging.

New Classification: "Street Server"?!

Caught this on a walk through San Francisco yesterday.

There is something of a tradition in the City to place unwanted furniture out on the sidewalk, hoping that someone will find use in it. It generally works for everything other than IKEA case goods, which usually don't make it structurally from the move to the street.

Well, what do we have here, leaning up against the lamp post, behind a chair?

Why, it's a server! I was pretty surprised, as I have often been told that there is a solid recycle/reuse industry in place that takes old data center and telco equipment back. 

In fact, in California, there is a consumer electronics recycling program in effect that pays a redemption amount for PCs, printers, televisions, and the like, which is why you don't see too much gear on the sidewalk (there are collection events where you can drop off the equipment for free).

Just a note from the utility energy efficiency program perspective: several key IT efficiency measures result in equipment replacement, and utilities are loathe to think that their programs contribute to improper disposal.

Utilities are also leery of "trade-in" incentives (i.e. paying customers to turn in outdated monitors when they buy efficient LCD units), because they fear paying for equipment that has already been relegated to the dark corners of garages or IT graveyards in company storerooms.

Let's hope that this unit was replaced by a new, premium efficiency model at least. Doesn't look like it was decommissioned as part ov a virtualization project, or there would be a stack of them on this corner!

Intel vs. ARM from The Economist

I've posted before about the potential use of ARM-designed chips in servers as an energy efficiency play. This article from The Economist focuses on the potential market battle between Intel and the ARM cluster of companies, noting that energy efficiency is the basis of the competition.

In a nutshell, ARM chips dominate mobile electronics applications, which of course means that energy efficiency is a prime design attribute (to maximize battery life). Intel dominates in desktops, laptops, and data center gear, and their processors bring awesome computing capability to bear at increasing energy efficiencies. (Disclosure: Intel is one of my clients.)

It's been posited that ARM-like designs would be applicable to Intel's markets on the basis of energy efficiency - a server running ARM chip sets could use as little as a tenth of the power of an Intel-based model, according to HP (HP has a project to develop ARM-based servers called "Moonshot"). ARM chips might also be suitable for PCs - certainly in the laptop, tablet, and new "ultrabook" categories.

The article points out many of the barriers that would prevent one side from gaining traction in the other's turf: the PC market is stagnant (so why bother), Intel has developed the Atom processor which has efficiency on a par with comparable ARM products, ARM designs likely lack the performance needed for server workloads, and the whole issue of whether software companies and other industry players will support new platforms.

I'll raise just two issues.

First, it has always appeared to me that issue of server energy efficiency is mooted by the incredibly low average utilization rates in most applications. Even the best internet-focused companies with largely homogeneous workloads only report rates in the twenty percent range, and enterprise users are lucky to see rates over ten percent. So efficient power supplies or even comparative advantages in processor efficiency simply don't make the difference that improved utilization would. 

Second, the PC market is relatively stagnant, but there are tremendous opportunities to drive efficiency through refresh and power management. The latest generation of Intel processors are far more efficient than cores made just three years ago, and coupled with power management, desktops use just a little more energy than a late generation laptop.

Which is why I counsel utilities to promote server virtualization and consolidation and desktop power management as their primary programs in these markets. Yes, there are discussions at some utilities about designing programs that reward the purchase of premium efficiency desktops and servers, but that will drive marginal energy savings compared to these two measures.

As to the ARM/Intel foray - battling over energy efficiency is a great thing!

 

Intel Site: Great Utility Data Center & IT EE Program Resource

This is one of those things I should have posted a long time ago, but a recent client inquiry at least prodded me to do it now.

The client wanted to know how to find out about utitlity energy effciency program offerings for data centers across the US, without having to do endless internet searches on his own.

Intel has created just such a resource at their Premier IT Professionals site, which features not only a map and lookup feature for program information, but a blog, technical resources, and useful links.

Full disclosure: Intel is one of my clients, and I have been producing some blog content for this site. But I don't at all mind you bookmarking it along with my blog!

ESource Forum Presentation Deck

I had the pleasure of attending the ESource Forum in Denver this week, sharing a presentation slot with Geoff Overland of the Wisconsin Energy Conservation Corporation to talk about how to prepare and launch utility energy efficiency programs for the data center and information technology market.

We got good feedback from several of the participants, and I like to think that we inspired some utilities to at least get in the shallow end of the pool.  We heard from a few utility program managers that they'd  like to start holding customer education and training sessions, which I advocate as a key component of a phase one program roll out.

ESource will be posting the presentations within the next few weeks, but here's my slide deck: Download ESource 9 21 11. I'd call your attention to the last two slides before the close, which summarize a two-phased approach to a program launch.

Thanks to the ESource team for having me on the agenda - I enjoyed being in Denver for the first time.

In Development: Multi-Utility EnergyStar Server Rebate Program

I've been asked by a client to try to put together a working group of utilities, IT equipment distributors, and program managers to offer a rebate program for EnergyStar servers.

Nominally the effort will look very much like the program for EnergyStar desktops and monitors offered by five utilities on the west coast, including three in California (PG&E, SMUD, and SDG&E). That program is moving something less than a million dollars a year in incentives, but by all accounts has the strong support of the participating distributors (after a period of some skepticism).

There are a number of roadblocks on the path to success, including the fact that server makers have not yet made the effort to certify many, if any, of their products.

I'm building a stakeholders list now, with the intent of holding a program planning "summit" in Q3. If you're interested or would like to participate, let me know.