Slashdot Mirror
Shortage of Electricity Drives Data Center Talks
Engineer-Poet writes "Per the San Jose Mercury News, competitors such as Google and Yahoo are meeting to discuss the issue of electricity in Silicon Valley. How much of the USA's 4038 billion kWh/year goes into data centers? Enough to make a difference. Data centers are moving out of California to spread the load and avoid a single-point-of-failure scenario. This is a serious matter; as Andrew Karsner (assistant secretary of energy efficiency and renewable energy for the Department of Energy) asked, 'What happens to national productivity when Google goes down for 72 hours?' I'm sure nobody wants to know." From the article: "Concern about electricity pricing and volatility has led Microsoft to talk with its network manufacturers about building more efficient servers. IBM and Hewlett-Packard -- which both build data centers -- want to improve efficiency at the facilities. AMD promotes changing the design of data centers to increase airflow to keep the supercomputers cool."
We just started switching from Intel to AMD hardware in our servers (HP DL385). Not that we pay per Kw/h, but I figure less power consumption means less heat and less fried hardware.
AMD has a website on the topic: Real Efficiency in the Data Center
"No matter where you go, there you are." -- Buckaroo Banzai
Now, I hope, people will start to understand why Sun and Intel are focussing so hard on performance-per-watt, and not just performance.
I am TheRaven on Soylent News
It's still a good idea to reduce server power because it reduces both the operating power AND the cooling power required.
On another note, has anyone noticed that language used impacts performance per Watt?
I recall from a /. comment from the son of a congressman(Congressional brat?), where he said that his father worked rather more than six days a week, due to the necessary reading of bills in his committee, and such.
Ninjas and pirates. How piquant.
People who manage and run data centers have to think it through before making changes. Many servers that are more than a year old were not designed for energy efficiencies. To top that, they weren't designed to take advantage of natural efficiencies in telecomms data centers. Most telecomms equipment is designed to run off of -48VDC. This has the effect of reducing the number of wasteful 115VAC to DC conversions along with the subsequent losses to heat that have to be removed by A/C systems. I've seen estimates that show the possibility of up to a 35% reduction in power and A/C costs simply by converting the AC power supplies in servers to DC power supplies.
Additionally, much of the forced air (from the floor upwards) A/C systems I've seen in data centers is not configured properly. There are vented tiles in places they shouldn't be, and not where they should be... causing hotspots and A/C problems in general.
I see datacenters with a wide variety of rack types. This can work, but often leads to inefficient use of the A/C systems. Its expensive to change racks, if its even possible (some vendors don't like their kit in someone else's rack) but this problem also needs to be looked at. A/C accounts for a huge energy drain in datacenters.
Using older hardware rather than buying new hardware saves on the short term, but the savings in energy costs by buying newer, more efficient hardware is something that datacenter managers HAVE to look at if this problem is to be solved. Its not just a matter of being 'green'. Its a matter of saving money that can then be used to bolster other parts/systems of the company.
I think that we'll see Google et al running VM clusters soon, where unused servers in the cluster are shutdown till they are needed for heavier traffic. In much the same way that complex automotive engines shut off several cylinders during low power requirement times, servers can be shut down (sleep mode) to save power until they are needed.
These are just some of the ideas that are currently the talk of datacenter managers and the vendors who support them. Try perusing the APC website, or other datacenter vendor's websites.
Support NYCountryLawyer RIAA vs People
There's no reason they couldn't just setup solar panel "power plants" in Arizona and add that power into the grid. I see an old dated power grid as a big concern though.
Another interesting tidbit for comparison: a typical high-density rack puts out something in the neighborhood of 15KW of heat. An average home electric oven puts out about 7-8KW of heat. So each high-density rack is like having two ovens going full blast, 24x7.
I thought i read that google now has power plants and data centers, in oregon, that are based primarily on water-power from a huge river.
So, if power goes down, they actually might be the only ones who *ARE* up and running, which is pretty fucking cool.
B
This is done in a number of gyms. The power produced by an average person on an exercise bike isn't enormous, though, and is used to power the display on the bike itself. There may be a tiny bit left over, but not enough to be any real use.
Data centers need to figure out a way to use the "waste" heat and turn it back into something useful, namely electricity. The problem is they generate a lot of heat, but it isn't hot enough, which seems screwy but for co generation you want as hot as possible. So the tech that needs to be developed (along with the obvious not generating so much waste heat through efficiency gains), is to find better ways to accumulate/move and use the low temp stuff they do have lots of. There are some alternative energy projects out there doing that with solar thermal to stirling engines for example.
Google should convert their huge parking lots into solar 'car ports' providing multiple benefits:
whoever designs/makes solar panels strong enough to weather car traffic will make millions. i had this idea screw parking lots. turn the roads into solar collectors. (not sure if anyone else has had this idea or not, but it occurred to me one time driving thru middle-of-nowhere, utah, that someone, at some time, was out there with an asphalt crew, laying this freeway for years on end. i'd be interested in talking with anyone who's interested in this, as it makes for interesting speculative fiction/campfire talk/etc.
Why not just move them to Alaska? You cool the building using ambient air from the outside passed through a multi-staged filter. Being that it's so damn cold outside, you don't need to run any compressors.
Life is not for the lazy.
Given US laws these days I suspect companies would prefer their data OUTSIDE of US jurisdiction....
That aside latency is not really a distance issue - its a network design issue.
If you put a big trunk of fiber (as my original comment was saying) from iceland to NY and iceland to london (thus making a nice redundant triangle with the current transatlantic connections) and connected it the the existing back bones sensibly the extra distance would not really be noticed.
Hops add far more latency than distance, most of the hops being to get up from your home/office connection to the back bone, then back down from the back bone to the company - hence why data centers are built where its very easy to get trunk connections - the proximity is a cost issue - fiber is expensive to lay.
$_="Slashdotter";$syn="OTT";s;..;;;sub _{print shift||$_};s!ash!Perl !;s=$syn=ack=i;tr+LLEd+BLAH+;_"Just Another ";_
Some data centers actually cool their facilities with air pumped in from outside their buildings. There's a study underway at Lawrence Berkeley National Laboratory looking at the use of air economizers at seven data centers that have participated in a PG&E program offering rebates for folks who do this. The study is looking at concerns that the use of outside air will introduce contaminants or excess humidity into the data center. Not for everyone, but seems to work for some folks.
RichM
Data Center Knowledge
I remember a few years back when we had a blackout here in Denmark. A datacenter a friend of mine was working for was somewhat proud that they managed to stay up during the blackout - only thing is, none of their peers stayed up, so they where not actually connected to anything :)
;) I dunno, I guess it is a sick kind of pleasure :P
;) No fan noise, no A/C noise, no noise from the plant...
But that is one of the cool things about an outage; okay I am a Sysadmin, so my job is to keep everything up and running... however during our annual plant shutdowns, I enjoy watching server utilization and network utilization drop to zero
Nothing like also sitting in our plant server farm rooms, listening to the disks spin down - A/C shutting off and then quietness!
Tes
If they had 3 such separate centers (which I have no doubt they already have), the only way they're going to be totally off line is if the whole national grid goes down - in which case Google should be the least of your worries if you're a lawmaker.
Especially if they have one in Dallas (or any large city in TX other than El Paso). The TX grid is the most independent of all electric grids, and rarely do problems traverse its boundaries.
Learn to love Alaska
Why not just move them to Alaska? You cool the building using ambient air from the outside passed through a multi-staged filter. Being that it's so damn cold outside, you don't need to run any compressors.
You might need to heat the air a bit - there are minimum operating temperatures for most gear!
Since most of the network traffic endpoints aren't anywhere close to Alaska, the additional 10-50 ms of latency might be a deal-breaker.
What does big bandwidth cost in Alaska? I have no idea.
I run a rather large department for one of the Investment Banks, with users / developers / support staff dispersed between London / Amsterdam / Cairo / Milan and Rio.
About one year ago the folks maintaining our applications infrastructure were advised by the companies responsible for the municipal grid to reduce our hardware footprint in London.
The reason? The grid was close to if not already overloaded, and increases in consumption were to be discouraged.
So we've been putting all new build into Central Europe, and slowly migrating existing systems over as we can.
A strange situation all around, if you ask me.
A message from our sponsor
What's the point of locating your datacenter in an area with high ground prices, a history of electric power supply problems and a hot climate? Usually, the reason for this is very simple: that's where your CEO's house is. If executives made business decisions based only on economic considerations, there wouldn't be any nationwide businesses headquartered in California, would there?
If slashdot went down, perhaps the vast army of users, presumably GNU/Linux geeks, might increase their productivity 100 fold.
...."Monkey Boy" o "Money Boy"?
Such a scenario suggests that slashdot is a Microsoft troll, luring the GNU/Linux community into "daytime soaps" (OK, evening and late night soaps too) instead of being productive coders and advocates to the larger community.
The counter argument is that slashdot readers and posters are lurkers only, and couldn't contribute a useful line of code anyway.
-----------
Since this is undecidable without benefit of an experiment (so where IS the main power switch?) perhaps the best thing to do is to set up a bizarro-slashdot for the MS "community" (yeah, make it one of those "I.E. only" sites!) and watch MS lovers fight over
"quien es mas macho"
Is it mere coincidence that Monkey en espanol is mono?