Slashdot Mirror
Low Energy Supercomputing
Faith Singer at TACC writes "The term 'supercomputing' usually evokes images of large, expensive computer systems that calculate unfathomable algorithms and run on enough energy to support a small city. Now, imagine a supercomputer, but run on the electrical equivalent of three standard-size coffee-makers. This year's international supercomputing conference, SC10, will feature the Student Cluster Competition that challenges students to build, maintain, and run the most-cutting edge, commercially available high-performance computing (HPC) architectures on just 26 amps."
Amps = current, not energy....
Can I use as many volts as I'd like?
I went to eat some animal crackers and the box said, "Do not eat if seal is broken." I opened the box and sure enough..
Power is now called energy, and is measured in amps? No one told me...
I don't know about computers, but you can get a lot of productivity out of humans with the power produced from three coffeemakers.
Gamingmuseum.com: Give your 3D accelerator a rest.
Try Joules (in context as a total), or watts (as a measure per unit time).
"The competition challenges students to build, maintain, and run a cutting edge, commercially available HPC architecture on just 26 amps of energy."
Only problem is that the Ampere is a unit of CURRENT, not energy. It's like saying someone weighs 686 Newtons.
While I understand that if the voltage is kept the same, then the amps are proportional to the energy involved per unit time because W = V x A. However 26 amps at 120 volts for 1 second is not the same energy as 26 amps at 5 million volts for 20 years.
Seven puppies were harmed during the making of this post.
26 amps at 350kV?
Why, without your clothes, you're naked, Miss Dudley!
Current * voltage = power. In the United States, alternating current from the wall is nominally 115 volts, and 115 V * 26 A = 2990 W. So I think the actual figure was supposed to be 3 kW of power. Run this for one eight-hour day* for 24 kJ of energy per session.
* This can be business hours (if interactive) or the most efficient cooling hours (if batch).
The server applications have to run in JVMs.
Gamingmuseum.com: Give your 3D accelerator a rest.
But who thinks theyll be using atoms? I remember reading that Microsoft, and others were looking into clustering them. The current atoms are more powerful than a P4 and use only 13 watts, 18 at 100% cpu.
At 18 watts each and 26 amps at 110 watts thats about 158 atoms.
158 x $100 each = $15800
2.86 kw x 24 x 30 = 2059 kwh
$0.20 (the outrageous price I pay for electricity) x 2059 = $411 a month to run this thing
$0.04 (the price I wish I payed for electricity) x 2059 = $82 (Now thats reasonable!)
--free small codec pack
this is a false comparison.
Just because I can hook a shark from a boat, I do no offer to wrestle it in the water.
unfathomable algorithms
Like the kind they use in ocean tidal research or sea bed topography?
Thank you, thank you! I'll be here all week. Try the buffet.
Trolling is a art,
Amps are units of current.
Volts are units of potential energy.
Watts are units of power.
Joules are units of energy.
I expect better from Slashdot editors; this is absolutely fundamental knowledge for a geek.
I don't fear computers, I fear the lack of them. -I. Asimov
This is a student project so the correct unit of energy is a "Library of Congress Stacked with Red Bull Instead of Books."
Now, you may convert that into Joules, if you care to.
Schroedinger's Brexit: The UK is both in and out of the EU at the same time!
And you have might have a typical datacenter...
For example, can they run their allotted current (at nominal AC 120 voltage, one assumes) for a few days to charge a capacitor and/or supercool the equipment in advance of the test?
This doesn't sound too difficult. The number one power-consumer is cooling. Distributing the same code over a larger surface area would allow you to reduce just how sophisticated and power-hungry your cooling needs to be. Any SIMD code will distribute just fine over such an architecture. If you're really clever, you'd design the cluster as a series of pentagons and hexagons, allowing you to build a geodesic. This would not only maximize the surface area but would also minimize the distance network traffic has to travel, networking being the biggest cause of latency in supercomputing. The really really clever geeks would then set up additional "regional" networks to allow for much higher performance when handling code that needed to talk much more locally, then distribute the code according to those regions. (Essentially, you then have a cluster of clusters.)
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
The computational hardware (processors, switch, storage, etc.) must fit into a single rack. All components associated with the system, and access to it, must be powered through the two 120-volt, 20-amp circuits, (each with a soft limit of 13 amps) for a total of 26 amps, provided by the conference. Power to each system will be provided via metered power distribution units The equipment rack must be able to physically hold these metering power strips.
This makes it even harder since theyir hardware has to be power balanced between the two power strips. They'll have to come up with some dynamic load balancing between cluster nodes based on power consumption. I guess dual power supplies might help (do dual power supplies draw perfectly balanced power between both power inputs?), but at a loss of power efficiency.
*Post complaining about amps not being a unit of energy*
*Feeling smug*
*Actually reading thread, then feeling not so smart because 10 people have already mentioned it*
Insanity: voting in the same two parties over and over again and expecting different results
Let me say that I really, really like this sort of a challenge. 26 Amps at 120V would definitely trip my home's circuit breaker, but it's a certainly an amount of power that's available to mere mortals. I have a feeling that we'll be surprised by all the creative ways the contestants will find to save watts. Since there is no size/clumsiness limit, I have a feeling that the cooling will be done by water pumped from a giant, passively cooling reservoir. (Well, there was a vague mention of "racks" - so maybe that won't fly.) But I genuinely wonder what sort of architecture will have the advantage here. My money is on ARM. Anyway, what's cool is that the result will be an awesome machine which could actually run in your house, and that's pretty sweet.
I've always found electric heaters (including geysers, etc. but mostly environmental heating) a huge waste of low entropy. You can achieve the same goal by powering enough chips -- would work especially well for floor heating. Now, if you're not recycling old computers, it might cost some, but if our only constraint is energy, we can thus create a supercomputer that spends 0 energy "for itself", just by installing this system to a few buildings.
You could even communicate through the power line, thus eliminating the need for a separate network installation. "Buy our @home geyser, that pays for itself!", that sort of thing...
You'll notice they specified two 120v outlets of 13A each. You could easily run that on two standard home outlets, as long as they were on different circuits.
Aside from his obvious misuse of amperage as an energy unit. Does any of this really make sense, why save power...crank that baby up.
Unit confusion aside, it wouldn't take much to be like "Hey, it doesn't use much power; let's get hundreds/thousands of these things working. Just imagine the unfathomable algorithms we can calculate!" Aaaaaand.... we're back where we started.
Imagine a Beowulf cluster of these!
sorry, been waiting over a decade to say that...
...aye, 'tis been done before, or at least (really) low powered boxes have been clustered e.g.
PicoCluster
Plan 9 Cluster
and all the way up to SeaMicro's 512 Atom beast
So get some more modern (than the first two examples) SBC's, put them into a rack form factor case (as per the rules), chuck in some Coreboot and then profit.
then this means on a 120 volt system, we're talking 3,210 watts to run the thing. That's about 3 kilowatts.
- Zav - Imagine a Beowulf cluster of insensitive clods...
Seems to me that a three-way SLI Nvidia config running CUDA would be hard to beat.
Do they write anywhere how much of the superconducting architecture they have? Number of cores? Flops? Mips? Anything?
Heck, i can build any architecture with a few watts if i am allowed to underclock and only use 4 cores to demonstrate it.
But well.... It was in Texas... Down where hillbilly creationists roam in the educational boards. Probably they where thought in school to strip any experiment of units and sensible numbers and replace these but general claims, so that all theories are equally scientific.
If it's that small and uses so little energy, is it actually worth the name super computing? Make a cluster of 100 of those, maybe then it's super computing...
Most switching mode power supplies work more efficiently at the higher AC voltage of 240 volts. So, what draws 26 amps at 120 volts should typically draw less than 13 amps at 240 volts. You can thus run more than twice as much computer capacity on the same wiring gauge, which means additional savings in energy loss through the wiring feeding to the computers.
Most homes in the USA have 240 volts. If you are going to put a lot of computers in a room, wire up a 240 volt circuit for it. Just be sure to use double-pole switches (which the power supplies do). The 120 volt surge protectors won't work or won't be safe. Get some 240 volt ones with the IEC connectors (same as the PSU input) which should have the double pole switches.
now we need to go OSS in diesel cars
...Stony Brook University, and the piece that's missing from the summary is 26 amps@120v, (dual circuits, soft capped at draw of 13 each)
Links to more info from the conference: SC10 CC Page, rules, and app list.
The competition is harder than it sounds, you have to build a cluster from the ground up, fit it into the power requirement (which means stripping out redundancies among other things), strip down a distro (we used Debian as a starting point), get the apps optimized, and then run through the data sets. Your team needs to *understand* the apps, the OS, and the hardware in order to win. There are several people from various teams from past years who have moved on to doing their PhDs in comp sci based on work from this competition (At Carnegie Mellon, MIT, and UMich off the top of my head).
It's important too, in a few ways. For one I know I learned more about clusters the first day I started working on the team for this competition back in 2007 than I ever knew before. That knowledge has led to research fellowships, jobs, and knowledge of what I want to (biochemical modeling). It's an experiance that very few undergrads get, and I think that's a shame.
For the industry it's an important highlight of what can be done with a lot of dedication and a focus on wringing the most from your hardware and software. and in doing that we showcase a lot of work that people dont think about. For example our cluster last year ran off a single disk, plus a large ramdisk as scratch exported over QDR infiniband to the compute nodes. No, it's not new, but it was novel to a lot of people who dropped by our booth.
For another, the ASU team was the first time *I* and many others ever saw a windows cluster in the wild.
Competitions like this are important, they showcase technology and introduce it to undergrads early, with positive benefit!
"goodbye and hello, as always" ~Prince Corwin, from Zelazny's Amber series
The drink of choice at the competition tends to be Mountain Dew :-)
"goodbye and hello, as always" ~Prince Corwin, from Zelazny's Amber series
Wait, could you perhaps please explain it using a car-analogy instead? I think I might understand it better then..
Ok this is retarded. Majority of posts are confused about basic energy nomenclature: Amperage, Watts, and Joules? The achievement of building a cluster and having to know the inner workings of an OS etc is a remarkable feat to occur in a short timeperiod. Kudos. And yes I did say retarded. Retarded people don't care, why should we.
Ok, so i'm sure a lot of students learn a lot, but after reading how this contest is run it seems like total BS.
First, with a power budget that small, building a cluster isn't even worth it. A large SMP machine will dominate on any bandwidth limited benchmark. A machine with GPU's is going to dominate any compute bound application. An 8 socket machine (aka DL980G7) packed out with low power 6-8 core processors is going to provide 96-128 threads. Load it up with SSD's, and Tesla's and your probably going to have to take the case off and start pulling blowers and such to keep it under 3kw (it comes with 4.8-9.6Kw of power supply). The SSD's will lower the power consumption over the spinning drives, but the Tesla's are going to eat it right up. Your going to save a shitload of power over a system packed with infiniband/etc interfaces and switches.
Secondly, the off the shelf limitations seem to be totally retarded as the teams are basically assigned vendors/hardware. That is probably to avoid the thousands of smaller vendors that actually provide "unique" machines that have power/performance numbers that are often an order of magnitude better than the dell/hp/etc machines.
Finally, At 3kw the whole thing could probably be nearly passively cooled, with just a little hardware engineering. Ripping out all the unnecessary crap, and powering down everything that isn't active probably would allow a machine with a total system draw much greater than 3kw to stay under 3kw with some active power management. This kind of stuff is built into most of the recent blade servers and such. The blade controller throttles the blades to keep the total system draw below what can be provided by the power supplies (aka the blades could draw 10kw, but there may only be 6kw of power, everything is fine as long as the blades operate normally and only draw 4kw, but when some unusual activity takes place the CPU's/etc are powered down). Again the problem is that the big vendors, don't give out this information, so your left at their mercy.
Lastly, again, making a "supercomputer" has gotten within the reach of real people lately. The Asus P6T6 WS has 6 PCIe16x G2 slots. A half dozen of these motherboards, with infiniband adapters, and GPU's could well put you in the top 20 of the top 500 supercomputer (due to the use of linpack) list for less than a full size SUV.
The Library of Congress has 1,199km of shelf space. (http://en.wikipedia.org/wiki/Library_of_congress)
A Red Bull can is 52mm in diameter (http://en.wikipedia.org/wiki/Beverage_can)
A 250ml Red Bull can contains 481kJ of energy (http://nutritiondata.self.com/facts/beverages/7399/2)
Doing the sums, I get:
Assuming a single row of cans: 23,057,692 cans, and 1.109x10^13 Joules.
If we have a nominal shelf depth of, say, 250mm - let's call it 5 rows of cans - we get 115,288,461 cans, and 5.545x10^13 Joules.
For comparison, a kiloton of TNT releases 4.184 terajoules, so a Library of Congress of Red Bull contains the equivalent of about 13.25 kilotons of TNT - about 1 Hiroshima Bomb, give or take.
Thank you. I realized that the next day when I checked my replies. It's 24 kWh or 86.4 MJ per eight-hour session, with three sessions sharing time on the cluster if cooling is available.
I fully agree. I was really hoping to see an article where hardware students are doing things like assembling ARM SoCs together into custom clusters, deciding tradeoffs between component count/speed and power consumption.
It's like they did their best to strip out any sort of true innovation from this contest, and company visions like SiCortex (RIP) are blotted from "the right way" to do things.