Slashdot Mirror
Is a "Net Zero" Data Center Possible?
miller60 writes "HP Labs is developing a concept for a 'net zero' data center — a facility that combines on-site solar power, fresh air cooling and advanced workload scheduling to operate with no net energy from the utility grid. HP is testing its ideas in a small data center in Palo Alto with a 134kW solar array and four ProLiant servers. The proof-of-concept confronts challenges often seen in solar implementations, including the array's modest capacity and a limited window of generation hours – namely, when the sun shines. HP's approach focuses on boosting server utilization, juggling critical and non-critical loads, and making the most of every hour of solar generation. Can this concept work at scale?"
Four servers is a nerd's basement.
Wouldn't you need something like 4 racks full of servers? Running something like seti@home or distributed.net?
In its own building.
As a net 0, No. It can't work from solar. The amount of electrical storage would make it impracticable.
However, this is a good idea, not as a net 0, but for cost and sustainability. Having solar during the day would reduce cost and cut down the backup generator requirement. If there is a brown-out/black-out on the power grid, during the day, you have solar. At night you'd still need diesel.
Fight Spammers!
That thought this "Net Zero"?
http://www.netzero.net/
Why would my datacenter want freaking banner ads all over it?
At the equator... then you'd have some uptime!
They're actually building it into a genius bitcoin mining mega-rig! No overhead for utilities?! They're rich! Okay, just kidding...although I didn't see any evidence that they're not, lol.
P.S. my electric costs were $40/mo and my bitcoin income from it was around $54 so yeah....but FPGA miners can run at 14W and can alone hit 0.63x the performance of my own Radeon card rig (which ran at around 480W). So setting aside the bitcoin mining joke, no matter what they're using the place for, eliminating the utilities is HUGE! It easily blows away hardware cost divided by useable life for certain server types.
Just build a nuclear reactor in the same building complex and then build your datacenter around it.
If you're already assuming that a data center can include its own power generation systems (solar, wind, hamsters, etc.), then of course it's possible.
Just include a local coal or nuclear plant on the datacenter's property. Or, if the "renewable" detail is critical, create one in the middle of the Mojave dessert, with 30 sq. miles of solar panels, which during the sunny times also charge up a 400-ton array of lithion-ion batteries or a flywheel generator.
So I wonder if "possible" is really what you're asking.
http://news.cnet.com/8301-13579_3-57436553-37/apples-main-data-center-to-go-fully-renewable-this-year/
They should build them in Nunavut.A couple of months are only daytime and it's cold enough that you'd be more likely to pay for heating to keep the machines within their optimum temperature.
A good $250K worth of solar power gear and that's if you're a bargain shopper. I'd estimate it will realistically generate almost 500KWH/day. That gives you ~20KW continuous load capability with net metering to remain net zero. I'd be surprised if those four servers at full load on all cores required even 4KW total, maybe half that typically so 48KWH/day. We don't get specs on their HVAC but in that relatively mild climate I'd say 50KWH/day would easily cool 3000SF with a modern heat pump in cooling mode.
You can't build something that will do the same job but not economically. These will remain concepts and prototypes until we can get a solar cell that is very efficient at a competitive price.
I wish this weren't the case... who likes being a slave to the grid but no one is making solar sustainable without absurd subsidies. The germans are making a big push for it right now which can only go to sad places because germany isn't known for it's sun and has huge energy needs. I wish them well and I hope I'm wrong... but it's looking to implode as soon as the maintenance costs start ramping up. We know a little about this in california. We've been building these sorts of power plants for decades. They work fine initially. But five years down the road everything goes pear shaped and you have another eco ruin in the desert.
I've decided to stop wasting my time responding to AC trolls/sockpuppets... so if you want a response from me... login.
Really, this idea has lots of warm fuzzy, but it won't work as a general industry practice.
Take my current location: last summer we had epic heatwaves of triple digit ambient daytime temps. Using open air cooling those servers would be actively overheating just by being turned on, let alone running code. Generating their own heat, they would risk serious failures. In palo alto, where they have a cooling ocean breeze blowing inland and moderating summer temperatures, it might work. In the landlocked hellhole I live in, average summer hatwaves would devistate their servers horribly.
Then you have the problem with the realestate needed for the solar arrays themselves and with the low efficiency of current energy storage technologies, and the high voltages needed to efficiently power a data center. Solar PV is DC, and is not made to supply high voltages. To get the high voltages, you need to turn it to AC, so you can use a transformer. Then you have to turn it back into DC at the high voltage, so you can efficiently distribute it through the datacenter. That's 2 great big energy vampires right there. Add the weaksauce of current solar generation efficiencies, and you have a "are you smoking crack!?" Moment.
While you might be able to utilize idealized conditions at idealized locations to make a low cost data center, you don't get to dictate that more than half the time. Somebody in rekjavik could make use all the geothermal power, and freezing cold outside for a really cheap datacenter too. Doesn't mean somebody in say, bahrain could do the same thing. Its not location agnostic.
Likewise, the power content of solar rays is not constant across the planet, so the costs of powering with solar vs the land utilization costs needed to run a datacenter that does more than just sit there to look pretty are non-trivial, and may well be outright impossible in certain areas.
I don't see this being more than just a pipedream.
No this concept can't work at scale - 130KW means around 10,000 sq ft of solar panels -- all to power 4 blade servers. If they were in a c3000 enclosure, they could put 8 blades in 6U - so could fit 56 blades in a 42U rack.
If they need 130KW and 10,000 sq feet of solar panels to power 4 blades, they'd need 14 times more panels to power a 56 blades in a full rack, or 140,000 sq feet of panels, all to power 6 sq feet of servers.
So a small 12,000 sq foot datacenter can hold around 1000 times more servers, so you'd need 140M sq feet of panels, or around 5 square miles of solar panels.
That doesn't seem very scalable.
... Then the answer is probably no. I used to stack Dells floor to ceiling in the racks and never had a problem with power. Just interleave a PDU every so often and plug 'em all in.
Then I got a job at an HP shop. Started putting DL360s and DL380s in a rack. Breaker pops. Break out the clamp meter. No, the breaker's no defective. Those things GUZZLED. I have no idea what they did with the extra juice.
Anyway, if that's what they're using, they should forget about it. But perhaps their hardware has improved since then. People are paying more attention to power these days.
If it's in California even expensive photovoltaics and a huge number of batteries are probably going to come out to less per MW than the prices from the basket case you call an electricity industry.
Also with the small amount of power required (in generation terms) you could run the place from power generated by burning things like the waste products of sugar cane milling (bagasse burning co-generation plants has been around for decades), methane from a sewerage treatment plant (decades again) or others from a long list of small scale power plants that could still feed a huge data centre.
To me an exercise in running four servers on solar panels sounds like a 1990s high school project to show the kids what sort of stuff was being put in to power remote communications equipment in the 1980s.
Can this concept work at scale?"
Almost anything can be done if you don't care what it costs. What I don't see here (or in the similar Apple and Google announcements) is any indication of what their cost target is. Does anyone have any idea what their electricity costs need to be (or what the average datacenter revenue per megaWatt is)?
You can pass the excess power on to the grid according to the definitions they're using. As long is you've given more power to the grid than you've taken out, you're a winner.
Does this experience account for the solar panel manufacturing costs and their environmental footprint as well? Even the most optimistic studies admit it is not zero.
You don't understand this electricity grid thing, do you? It isn't "dumped on the grid". It gets used. While it supplies power to the grid, either other generation has a slightly smaller load or there is a minute voltage rise (which causes things like ovens to warm up a tiny bit faster so they reach desired temperature a little quicker.) This is the whole concept of the grid - lots of generation that comes on at different times and is carefully managed to meet the load requirements. Has it occurred to you, for instance, that when a conventional generator is down for maintenance it takes power from the grid to supply lighting, heating and equipment? We don't say it is useless because it has to be shut down periodically.
From scarped cliff or quarried stone she cries "A thousand types are gone, I care for nothing, no not one."
With the current management style, there will be so many reporting tools for middle management left and so few workers, that they'll need more than 4 servers per employee to fill in time sheets and surveys.
I was promised a flying car. Where is my flying car?
I haven't heard about that ISP for long time, no they are building a Data Center! It's amazing some of the things the things I don't hear about.
these guys http://www.netzero.net/
Sure it's possible to have a Net Zero datacenter: Just use unary coding instead of binary, with 0 as the unary digit.
For example, by binary number 101 would become as 00000.
See? As a result, all the bits will be zero. Net zero. That means the data will weigh less, and it's good for the environment! A win for everyone.
ground level solar will never work to power the cloud. there's too much shade. duh.
Serenity now, insanity later.
... Beowulf cluster of those. Secondly, first post. As in the first one that I've ever done.
There is no such thing as a net zero anything when you look at the dust-to-dust analysis. Things always cost energy.
In this case, it is only a net zero proposition when observed inside the confined bubble of the datacenter. However, when you look at the total energy expenditure, it is hopeless to try to call this a "green" solution.
Currently, it takes about 30MWh of energy to produce 1kW of full-sun solar capacity. So, their 134kW solar farm cost about 4GWh. Of course, with a 30,000 hour full-sun payback cycle, the solar farm will never produce that in its useful lifespan of 25 years, let alone with most of its energy being consumed by the datacenter.
This is not even to mention the energy cost of building the datacenter, and the things inside the datacenter, maintaining them, and then managing the waste when they are discarded.
So, no, this is not a net zero project - not by a long shot. In fact, I would venture a guess that it is actually more costly from an environmental standpoint than just building a traditional datacenter powered by electricity produced from NatGas or Nuclear, especially when adding the toxic waste emissions from producing silicon solar cells to the tremendous energy required.
The Republic of Colombia sits between Panama and Brazil. Its population is the second largest among Spanish-speaking countries.
We run a datacenter in a LEEDS Platinum building...
http://www.nrel.gov/news/features/feature_detail.cfm/feature_id=1505
I think the "net zero" is referring to HP employee pay. You work and net zero pay. Data center (or account) becomes profitable; as a former employee I have some first hand experience with this concept from HP.
1. Spend $X buying bonds
2. Use a portion of the interest to pay electricity bill.
3. Re-invest remainder.
For some value of X, you will earn enough in interest to pay your energy bills.
X will *certainly* be less than the amount of money required to build your own mini-power generation facility. I understand the need for on-site emergency backup, but in terms of day-to-day operation, your own boutique power plant will be more expensive than buying power at market price from the grid.
Economies of scale, people -- we do not need to make our own shoelaces, we can buy them from someone who specializes in doing this. Energy is a commodity, and HP shouldn't be investing in expensive ways to make cheap commodities.
When in doubt, have a man come through a door with a gun in his hand.
Usually that's what we call a guy with a CCNA and no real-world experience.
First off, the four servers are "fairly" beefy - four servers, each with dual 12-core CPUs, 64 Gig RAM, etc. and their datacenter include cooling (running air conditioning off of a solar arry seems like a bad idea, kinda like running a laser printer off battery power)...
If you really wanted to shoot for zero power from the grid, rather than throwing a huge solar Array at powering (and cooling) four big servers, why not re-engineer the datacenter to require less power all together? If we have to pack a dozen or so high-velocity fans inside a 1U rack server, maybe we should consider sacrificing space for power demands and use a 4U case that can tolerate higher temperatures? Why not investigate powering the datacenter off of DC, rather than several conversions to and from AC to DC? Why not study the concept of more lower-power CPUs instead of fewer high-power CPUs?
All they are really proving is that they can run a datacenter with a power requirement of NKW by building a solar array that supplies NKW x 24 hours worth of power every day in the few hours each day sunlight is available each day.
Maybe I should embark on a similar effort, put an HP microserver in a tool shed out back, throw up a few solar panels and deploy some batteries, then have my cable company run a line out to the shed and have my server & ISP router powered off the batteries charged by the solar panels... The microserver doesn't need special cooling, and it requires just under 200 watts electricity, so figuring in a small safety margin of 50% to power the cable router and loss due to storage/conversion I would need 24 x 300 watts about 2.25KW solar array and I figure a 24 hour battery that could hold 7.2KW of power (what is the proper scale for such an item? Half the daily power consumption?)
Only the 2.25KW solar array would be about 224 square feet - to power one little HP microserver 24x7 (or any 300 Watt load).
Ken
Where the rubber meets the road is: ? can solar (and similar energy sources) power the machines in Congo that dig the coltan out of the ground for the warlords to sell to ICT companies, and can solar (and similar energy sources) power the machines that build the ICT machines in the data center, including smelting the aluminum out of the bauxite, and the electrical power needed to pull the Al out of the molten rock and then refine the remains, and then can solar (and similar energy sources) power the machines that smelt and fry up all the other elements needed (silicon, indium, gallium, germanium, etc.) and then can solar (and similar energy sources) power the machines that deliver all of these components to the factory in China where all this crap is manufactured, and then can solar (and similar energy sources) power the machines in China that build the ICT and deliver the ICT to the data farms and homes and businesses that depend on this technology.
If yes, then industrialism can last for at least another century or two. If no, then no amount of wishful gee whiz thinking is going make it happen. Thermodynamics isn't just a good idea - it's the law.
Shoes for Industry. Shoes for the Dead.
It has always been "possible". Why it is not done every day is cost. The question really is can we afford net zero.
1kW of full sun solar capacity cost less than 2.000 USD in a large installation.
Let us assume solar farms and the equipment used in them are designed, built and marketed by little electric elves that solar farm operators call by sticking a small carrot in the lot where they want to erect the solar farm.
All the operators have to pay for is the electricity to operate the elves.
At 30MWh per kW peak installed power, the energy cost alone would be 3.000 USD per kW peak at a rate of 10 Cent/kWh for the electric power running the elves.
I am not sure where your 30MWh-figure came from, but it sure is baloney!
Even in aluminum smeltering, energy costs make up only about half of the total production costs.
The should of gotten Carly Fiorina to power the data center using all her hot air. They clearly missed the boat.
Many HP Enterprise Services employees work at home, using their own hardware. "Net Zero" for HP, for sure.
Net Zero green datacenter? Already done:
http://www.aiso.net/technology-network.html