Slashdot Mirror
Cirocco Live Liquid Cooled Rack
Mark Grant writes "Cirocco have developed a liquid cooled rack of AMD Duron 1.1Gs in a Beowulf cluster. The rack has been installed in Cambridge University, England and has been under trial since Christmas. The system is being put through its paces running chemical research algorithms.
Critical to Cirocco's liquid cooling system are the hot swappable quick couplings. These allow servers to be disconnected whilst the cooling system is in operation." The graph with live temperature readings is pretty neat.
Taking just one of those from the cluster.
it would be like have a pc.
wouldn't it
Can you imagine a Beowulf cluster of... DOH!
Viv
Gmail invites for ip
The idea of "Hot swappable" when it comes to cooling couplings is making my head spin.
You have to wonder if "hot swappable" is the right term for this kind of system.
Why can't I moderate something "Wrong" or at least "Grossly Misinformed"?
While it's possible to secure this sort of thing so that water doesn't leak out from the cooling system, it's much harder to manage a _very_ large cluster of processors cooled by water. Cooling a cluster with air is easier, because you only have to watch to see if a fan breaks down, and that can be automated. Cooling a cluster with water is more dangerous in that you really should inspect it visually every so often to see if moisture is leaking out from the couplings (or condensing on the pipes). So really, could you imagine having to maintain a Beowulf cluster of these?
didn't cray master the liquid cooled cabinet design, like, 30 years ago?
Personally, I always thought a liquid cooled rack is what happens when Pamela Anderson spills beer down her shirt...
Do those special, magic, fluid connectors look like scaled down versions of ordinary hydraulic dry-disconnect spools to anyone else?
I wonder what type of chemical research these systems will be conducting... perhaps they will determining the reaction between water (H2O) and Silicon printed circuit boards? (come to think of it, reasearch isn't the only thing they'll be conducting)
-- If you try to fail and succeed, which have you done? - Uli's moose
I'm missing something. As cool as it is, why do you need to liquid cool 1.1Ghz Athlons. Its nothing a fan can't handle adequitely and at a much more desireable cost. Are they just going for the wow factor, or is there an actually reason for the liquid cooling.
Was going to make a beowulf joke, but then you insensitive clods would mark me redundant (I'm only like the 20th poster, how redundant can I be?)
YOU SUCK BALLS!
For those of use who've grown used to making "Woo, imagine a Beowulf cluster of them!" jokes yet have no clue what a Beowulf cluster actually is, the definition, history and so on is available at:
NASA's Beowulf site
In brief overview:
In the summer of 1994 Thomas Sterling and Don Becker, working at CESDIS under the sponsorship of the ESS project, built a cluster computer consisting of 16 DX4 processors connected by channel bonded Ethernet. They called their machine Beowulf. The machine was an instant success and their idea of providing COTS (Commodity off the shelf) base systems to satisfy specific computational requirements quickly spread through NASA and into the academic and research communities. The development effort for this first machine quickly grew into a what we now call the Beowulf Project. Some of the major accomplishment of the Beowulf Project will be chronicled below, but a non-technical measure of success is the observation that researcher[s(sp)] within the High Performance Computer community are now referring to such machines as "Beowulf Class Cluster Computers." That is, Beowulf clusters are now recognized as genre within the HPC community./i
Perhaps they should look into using one of these for their webserver.
Even more impressive when slashdotted.
yeah I used to have a Scirocco, water cooled and everything. A great Volkswagen but damn what a parts hog. 2.8 Liter engine and a 5 speed manual it was a blast to drive.
Oh Cirocco? Not Scirocco? Whoops, but why on earth would anyone would dig up that dead convoluted name is beyond me. A few VW enthusiasts might always remember you but I think your just alienating your audience, all naming your company after trade winds.... Maybe iCirocco? Nah.
---- The real Slashdot is still here. You just have to browse at -1 to read the comments.
That they don't use their products on their own webservers...
/. effect nabs another victim!
The
This is a stupid idea. They say that air conditioning is inefficient, but they could have easily done it efficiently with ductwork.
I've worked with quick-couplings on megawatt lasers, and I can just give em one tip: couplings fail more often than computers. Just wait til they spring a leak because some idiot forgets to twist the ring properly, and he floods the whole rack.
You know what I'm talking about. Don't do it.
There is nothing so good that someone, somewhere, will not hate it.
You know, in a way I'm disappointed. I had always hoped to find out that Beowulf Clusters only acheived prominence after having slayed the evil Grendel Clusters in the back of some Scandanavian CompSci lab...
If brevity is the soul of wit, then how does one explain Twitter?
Just out of curiosity, does anyone know what job opportunities are like in computational chemistry?
hot swappable
quick couplings
Now I feel dirty. In a bad way.
can you imagine a whole Beowulf cluster of theese things ?! well...huh...nevermind :)
____
nico
Nico-Live
I wouldn't mind having one of those to get my juices flowing, if you know what I mean. Wink Wink, Nudge Nudge.
You see? You see? Your stupid minds! Stupid! Stupid!
Wow! Thanks for providing a link that's in the article. You're the best.
<tongue firmly in cheek>
I'll be gentle, seeing as you're clearly a newbie...
In time, when you've had a chance to read more heated arguments on Slashdot, you'll soon learn that we rarely take the time to actually read the articles. We just get on with ranting on them. I blame Hillary Rosen, the RIAA and Microsoft for that - in no particular order.
Now, if you'll just go to the second doorway down the hall, I hear there's an opening for a Grammar Nazi.
</tongue firmly in cheek>
Can you imagine a beowulf cluster of these ?
Oh wait.
Shit.
actually if you substitute natalie portman for pam anderson and hot grits for beer then this post has it all!
oh yeah, in soviet russia all of us are belong to your base.
If you mod me down the terrorists will have won
ahhh dang it ... just forget it!
Bet this
...that "Live Liquid Cooled Rack" was some sort of wet T-shirt contest for geeks ?
Seriously though, match this with the IBM Ice-Cube storage cluster and you really would have one cool machine (ducking).
"In a few years, one storage administrator should be able to manage a petabyte of storage, which is 100 times more than is typical today." - IBM Almaden Research Center
"Whoever would overthrow the liberty of a nation must begin by subduing the freeness of speech."--Benjamin Franklin
Well, although the current application is a stretch, the need for water cooling pops up when you have about 6kW in a cabinet. Air cooling is ineffective for data centers when you have more than this in a single rack. Getting the heat off the processors reduces your air-cooling requirement by 80-90%.
Also, with the water cooling in this type of application, you aren't as sensitive if your cooling equipment fails; you just need small pumps circulating water in the racks.
I am thinking something like a front-to-back fan design to blow the hot air out of the 1U cases (and I know those cases are pretty tight and cramped, so intelligent routing of cables, or custom device interconnects would be in order too) to the rear of the rack, and into an integrated (with the rear door) duct that funnels the hot air (with a fan) into the ceiling (or outside). Perhaps a better solution would involve heat pipes leading away from the CPUs to the duct (less fans).
I guess what I am trying to get at is that an intelligent design on the part of the 1U cases and the rack could probably acheive great results with air-cooling alone, and be cheaper for the end-user of the machine (from both an initial cost standpoint, as well as maintenance) as well.
I am not saying that liquid cooling doesn't have it's place, but this doesn't seem it (of course, I could be completely wrong, as I have never worked with such a dense system before - I do know server rooms can heat up quick, even when they aren't filled - so a dense rack very well might be a big issue and need liquid cooling)...
Reason is the Path to God - Anon
Hey, The temperature is increasing on the the cluster, probably as more people are checking out the live temperature monitor. Also, it seems to be loading slower.
You can literally watch the slashdot effect on a server this way.
That what was all this school was for... to teach us how to solve our own problems. -- janeowit
Why not just install rackmounting rails on the inside of a refrigerator? That's what I would do if I were stupid.
It already is one, a beowulf cluseer would be just the same but bigger. Move along.
Even though the heat capacity of gases is generally larger than liquids, the thermal conductivity of water is about 30 times larger than air. Also, plumbing lets me move the water to exactly where I want the cooling to take place without heating it along the way.
IBM mainframes (ECL-based) used water-cooled plates for the CPU and have spent a lot of design effort on quick-connect couplings that don't leak. I just wish they had transferred some of that knowledge to the Sears washing machine group.
The later Athlon XP cores (thoroughbred) run much cooler at higher clock speeds and faster. So downclocking such a chip would result in less cooling needed.
Imagine a Beowulf cluster of... damnit.
I just took a quick look at the site (at 7:09pm est one of the machines, clinux7, showed at 55C - wonder what happened) and it only shows 9 machines. Apart from that its a very neat idea if you consider the efficiency of water vs air cooling. Why chill a large room when all that you need to cool is a small chip. The trouble is one chiller and many small areas, solution is to have some way of disconecting an area (machine)if you have to do some kind of repair. I think that I would have use taps, and out all them all in parallel, but I'm not a plumber.
Semper ubi sub ubi
*runs from the flames*
In soviet russia, your sig says "Super lemons clone YOU!" And aren't the grits supposed to go in the pants? I don't think you can compare Portman's rack to Anderson's.
But seriously, go easy on these jokes. I found out the hard way that nobody but you and me find them funny anymore. My karma's been like a yo-yo lately. (here it goes again!)
P.S. You forgot the beow... never mind.
Ron Paul 2012
geuse they know not to put gellow in it next time
has someone been reading classic literature again? And then confusing sed literature with their own fantasies?
I think the answer is yes
YOU SUCK BALLS!
many good reasons to use liquid cooling. firstly, it's *very* efficient and it allows very high volumetric density. secondly, there are times when air cooling is a Truly Bad Idea, like on boats, ships, and submarines, but also even "ground-based" transportation applications. It's pretty clear that blowing salt air over a circuit board (even with conformal coating) is a Bad Idea(tm), but it's also true in cars, trucks (aka "lorries"), and things like earthmovers. and in large systems, the efficiency part is a Really Big Deal. Heating up a huge flow of air on its way to the ceiling, just so the air chiller can try to move it back down to get sucked in again is just stupid except at very small scale. the Beowulf cluster in question may not *demand* liquid cooling, but you don't have to build one a lot larger for the difference to matter a lot. (This is especially true in Europe where rooms are not airconditioned to the degree they are in the US in the first place.) rather than upgrade an entire AC system, which probably involves lots of work on a 500 year old building, just run the loop to a remote chiller and declare victory. as for "it will leak", note that almost everything electronic in a submarine is liquid-cooled. it is true that the primary loop through the cold plate may not be water, but it gets to a water loop pretty quickly. no-leak connectors have been around a very long time - just not cheap ones. a large demand for better cooling technology is important to drive down the costs and to make it commonplace and not just the province of "Big Iron" (supercomputers or otherwise). -mo
oh, wait...
~head explodes~
...if I petrified it and poured hot grits on it? Would the cooling system still work?
If cooling is a real problem, the usual solution in dense avionics racks is engineered airflow. All heat sources are measured, and small ducts and diverters are sized and built to deliver air to the key spots, while not wasting it on stuff that doesn't need it. If you're building rackmount servers as a product, it's worth the trouble to do this. Otherwise, most of your airflow is wasted.
CPU's: 37-44, Ambient: 24, water in: 26.5, water out: 26...
Isn't "cooling" supposed to create more than an accidental 0.5 degree temperature difference?
Caveat Emptor: this message won't selfdestruct if you memorize it!
I would expect that they had tested Duron vs. Athlon in advance, and bought the cpu type and number that gave them the most bang for the buck.
Suppose they only run stuff that can fit in the cache size of the Duron, then the extra price for the athlon is a bad idea.
Yup, a beowulf of beowulf clusters is an even more
buzzword compliant "Grid". The idea is that users of
cluster A can submit jobs to cluster B somewhere else in the world, if cluster B has free job slots.
There are some serious challanges with this, to do with authentication, and how you deal with jobs that need or produce lots of data. Moving it back and forwards over the grid can take more than doing the computation if you are not careful.
Take a look at the globus project which attempts to create a standard middleware between different sorts of cluster.
http://www.globus.org/
I don't own A Duron, let alone many Durons that would necessitate an entire rack.
"A coward dies a thousand deaths, the brave but one."
Busted!
If brevity is the soul of wit, then how does one explain Twitter?
Actually Grendel (the book) isn't classical at all is it?
Shows you what I remember from Schweitzer's class.
TTFN
Very few things actually get manufactured these days, because in an
infinitely large Universe, such as the one in which we live, most things one
could possibly imagine, and a lot of things one would rather not, grow
somewhere. A forest was discovered recently in which most of the trees grew
ratchet screwdrivers as fruit. The life cycle of the ratchet screwdriver is
quite interesting. Once picked it needs a dark dusty drawer in which it can
lie undisturbed for years. Then one night it suddenly hatches, discards its
outer skin that crumbles into dust, and emerges as a totally unidentifiable
little metal object with flanges at both ends and a sort of ridge and a hole
for a screw. This, when found, will get thrown away. No one knows what the
screwdriver is supposed to gain from this. Nature, in her infinite wisdom,
is presumably working on it.
- this post brought to you by the Automated Last Post Generator...