Data Centers Prepare for a Renewable Future

miller60 writes "A small but growing number of data centers are generating renewable energy at their facility, despite challenges with cost and scalability. In a special report, Data Center Knowledge looks at data centers implementing on-site solar power, wind energy, geothermal cooling and recycling waste heat from their hot aisles. Even as some projects choose to go green, other data center operators insist that improved power efficiency offers a far better return and carbon impact than pursuing on-site renewables."

Comments

Computing power per watt

[Yvan256]

Use the hardware that give you the most "computing units" (targeted to your computing needs, i.e. floating-point, database access, etc) per watt. That should automatically take care of not using wasteful (heat-producing) hardware.

Re:Computing power per watt

[girlintraining]

Use the hardware that give you the most "computing units" (targeted to your computing needs, i.e. floating-point, database access, etc) per watt. That should automatically take care of not using wasteful (heat-producing) hardware.

There are trade-offs there as well. -_-

Re:Computing power per watt

[skids]

I'm all for more efficient hardware systems. They definitely, definitely should be done, as long as the efficiency gain is more than incremental. However the danger with taking that approach only is that it encourages rampant expansion as software engineers hear stuff like "we can do twice as much per watt now" and think "goody, I can save several man hours by not compiling out the debug code and re-testing the optimized code."

Same goes (kinda) for virtualization. It's a great technology, very smart, and definitely should be done, but it needs to be kept in mind that in the majority of cases the core reason behind doing virtualization is that you need to run software that is too crummy to share a machine with other software. Virtualization's downside is it becomes an excellent excuse for not applying pressure to your software vendors to get their product up to snuff.

Recycling waste heat is a very good thing. Solar and wind need to be grid tied for IT use because they are intermittent. If the IT staff is actively involved in the energy system, it can offer an incentive to optimize systems and software -- the same way the instant-mileage meter on newer dashboards encourages leaner driving. However if the IT staff isn't invested in the idea they don't really care whether they can stay within the power budget. In fact in most cases, they never see the electricity bill themselves -- it isn't an IT budget item.

So the reasons to "go green" on a data-center level (as opposed to the whole office space) is purely psychological. Not that it isn't a good one, just it merely serves to psychologically invest the IT team.

hrm.

[Anonymous Coward]

"other data center operators insist that improved power efficiency offers a far better return and carbon impact that pursuing on-site renewables."

These are not mutually exclusive.

Re:hrm.

[eln]

Exactly. Any data center engineers running into issues with the cost of power and cooling (ie, all of them) should be exploring both of these options. Whatever solutions we come up with, be they efficient machines or renewable on-site energy or both, are destined to become inadequate at some point as our computing needs inevitably expand to fill the available capacity. The more additive solutions we can pile on, the longer the time horizon before we end up in another crisis where we have to find yet another innovative solution.

Re:hrm.

[Monkeedude1212]

While true, the limitting factors usually boil down to either apathy for the problem, not enough money to deploy all solutions, or the profit motive.

Re:hrm.

[Kjella]

are destined to become inadequate at some point as our computing needs inevitably expand to fill the available capacity.

Which is why you can question how "green" these projects really are, what's the point to reduce consumption if it's immediately offset by an equal and opposite increase in consumption? It's as if we made cars with twice the MPG and everybody decided "cool, then I can make my commute twice as long and really get out of the city". Green projects are those that reduce aggregate consumption, sure it's nice if people get more for each watt or more people are able to participate in the wealth, but it doesn't do the environment any good if the total stays the same or goes up.

Re:hrm.

[XanC]

Not, gasp, the profit motive! The only good datacenter is the one in business to NOT make money, that's what I always say.

This obseesion with efficiency

[countertrolling]

It makes for very delicate, frail, some times dangerous products. Stone age equipment running on renewables is much more robust on the long term.

id like to see

[nimbius]

exactly how much in the office, not the datacenter, we're failing to "go green." I know its off topic but aside from the lights-out datacenter not much has really been done for large datacenters like the one i work in, while the office seems like an energy hog with no end in sight

I do hear constantly however of minimum light levels that must be maintained in offices, and the temperature in a cube farm being forcibly maintained at 72 degrees. the vending machines run 24/7 when nobody is around, and the parking lot is constantly lit up like a runway.

Re:id like to see

[jridley]

I agree, it's pretty damned inefficient here. They build buildings with the cheapest construction techniques available, just basically a big metal semi-insulated crackerbox, then put 10,000 tons of A/C on the roof. There are 45,000 watts of fluorescent tubes just on this floor of our office building, and they're on for hours a day that are not necessary. I'd like to be able to turn my computer off, but every few days I need to access it remotely. I could do both if they'd deploy a little 50 watt wake-on-lan box, but nobody cares. I have asked about 10 times over the last 5 years for the "green team" to push for "put the monitors in power save mode after 30 minutes" to be the default on the standard desktop, but though they always say "that's a great idea" it never happens.

Re:id like to see

[icebraining]

Wake on Lan doesn't work remotely, because it works below the IP layer (a disconnected computer can't have an IP address, since the TCP/IP stack isn't running).

Re:id like to see

[icebraining]

And that 1 machine is the box he's talking about. With the difference that the "box" would use much less power.

Re:except of course, when you leave it idle.

[Yvan256]

Idle means zero computing units and running at 50% capacity means half the possible computing units (but probably way more than 50% power usage).

Is "Web 2.0" really necessary?

[Animats]

Maybe the way to cut energy consumption is to dump unnecessary "Web 2.0" junk. Serving static pages is very cheap. Is it really necessary to generate the pages on your site from some "content management system" which makes multiple database accesses just to display essentially the same page over and over?

Both?

[2obvious4u]

Even as some projects choose to go green, other data center operators insist that improved power efficiency offers a far better return and carbon impact that pursuing on-site renewables.

Why is it that everything must be polarized. Why aren't we doing both? Both are good and they are not exclusive.

Imagine

[codepunk]

Imagine the fuel savings if every office worker that was able could telecommute instead of burning fuel to get to a job that could be done from home. Imagine the money saved on road maintenance and other things associated with the reduction in traffic. For at least the last 15 years I have commuted back and forth every day to perform a job I could have done without ever leaving my home.

Re:Imagine

[cj_nologic]

Imagine the fuel savings if every office worker that was able could telecommute instead of burning fuel to get to a job that could be done from home. Imagine the money saved on road maintenance and other things associated with the reduction in traffic. For at least the last 15 years I have commuted back and forth every day to perform a job I could have done without ever leaving my home.

Most office workers would be incapable of working from home all the time - partly from lack of motivation, partly from lack of social interaction. A lot of face to face meetings are pretty important too, and much as they are laughed at, water fountain conversations are where a lot of personal and business relationships are fostered.

The real alternative would be if people lived close to their offices and walked or cycled in every day. It would improve fitness, but also improve communities. In my country lots of people cannot afford to live close to city centers or business parks where they work, so they live in large housing estates on the edge of the city. This leads to estates where essentially everyone is away all day, everyone spends an hour or two in their cars instead of socialising or spending time with families. Then people spend the weekends in their cars again because suddenly all their friends, shops, social and sporting engagements are long distances away from where they live, and so the estates themselves never achieve a sense of community. And, of course, lots of fuel gets burnt.

I'm not sure how this can be rectified - town planning, perhaps, but it also needs a cultural mind-shift away from the car as the preferred mode of transport. It amazes me to see how people will drive 10 miles to the sports center, cycle 20 miles on an exercise bike in an air-conditioned room under strip lights, then drive 10 miles back. WTF?

Re:Imagine

[JumpDrive]

This has been going on for a long time. The current problem is that nobody thinks you work when you are at home. Here we have people who actually come in and put their head on their desk, but they are deemed good employees because they show up.

Basically it boils down to management. Usually it is deemed that you have to visually see the people to know they are working, otherwise you would have to know what they are working on to determine whether they are working effectively.

Re:An Efficient Office

[MrSteve007]

While it seems like a shifty answer, it all depends on how you calculate return on investment . . .

Without any tricky accounting, the simple ROI of the PV array is about 12 years. If you calculate the ROI based on the PV and battery backup, factoring lack of downtime, in our case it was closer to an 18 month ROI. For us, each hour of downtime translates to roughly $5,000 of lost earnings. It doesn't take too many hours of backup power to pay off. Of course we could have gone with a gas generator for a much cheaper installation cost, but the tax benefits of a PV array for a corporation can be very attractive.

As for the virtualization, that was based more on the regular 5 year replacement cycle we have on our hardware. Instead of replacing 5 old servers with 5 new servers, it was much more simple and cost effective to build out one powerful machine and virtualize the existing machines.

The cost of the ductwork and fan was about $1,500 - however our A/C unit consumed roughly $1,000 a year to cool the space, so an 18 month ROI. Of course, it's all dependent on your climate, building layout, age of equipment and ease of installation. For us, it's worked out well. We've now achieved a 75% reduction over our baseline from 3 years ago in our grid energy needs, while increasing processing power, lighting levels, and maintaining a comfortable climate controlled office.

http://jbdg.com/results.html