Slashdot Mirror
Transmeta Meets Blades
The Griller writes "Gordon Bell, one of the creators of VAX, and Linus Torvalds were at the launch of a new supercomputing platform at the Los Alamos National Laboratory. Based on Crusoe processors from Transmeta and running a version of linux, it is aimed at being cheaper than conventional supercomputers by requiring no cooling and lower maintenance.
" Basically, it's blade clustering, using Beowulf.
Oh well, here's a list of mirrors...
I've got to wonder why they are using Crusoes. It's a good chip for the application, don't get me wrong... but the last I heard the main advantage it has over StrongARM is x86 compatibility, which shouldn't be an issue here.
=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Friends don't let friends enable ecmascript.
We can make a beowo... oh. How can I make a "beowolf cluster" joke when it IS a beowolf cluster? Wah!!!
Just Imagine a Blade cluster of Crusoe processors running Linux.
Ok, I'm sorry.
Dude, imagine a beowulf cluster of...*KRONK* [hercynium is clubbed with a shotgun and dragged away by the moderators...]
I'm done with sigs. Sigs are lame.
Imagine if these weren't clustered...
Given that you don't need to actively cool these chips, I think what would be even cooler(N.P.I.) is a cube of chips stuck together and interwoven with some sort of vascularized heat-sink. A meaty cluster of 100 chips you can hold in your hand, and plug into a big cube-shaped socket on your supercomputing motherboard. Now *that* would be New for Nerds.
Websurfing done right! StumbleUpon
Great news for the Linux community. Torvalds keeps on keepin' on. Bravo Linus.
It all comes down to "power consumption, size, reliability and ease of administration", apparently.
And the marketing people at RLX Technologies should be shot for not having a press release up for this, as it's all based on their product...
imagine a beo... ahw crap.
The only trick would be getting the things to work properly in a headless configuration -- Apple won't ship them without a graphics card, but I'm relatively certain that you could get a LinuxPPC installation to work even without the card installed.
and ease of administration
could someone explain how a microprocessor is administered?
Gordon Bell created Linus Torvalds? Or is he just a clone of the real Linus Torvalds?
A programmer is a machine for converting coffee into code.
Wouldn't it be great to have a Beowulf cluster of these? Oh, wait. Nevermind :)
Lemon curry?
i am not sure how they are making an advantage out of the code morphing, but the article states that it is a primary consideration.
I don't even remember one of us slashdotters responding to a Transmeta story with 'imagine a beowulf cluster of these...'
Kevin Fox
http://www.infoworld.com/articles/hn/xml/02/04/17/ 020417hndoe.xml A link to the real supercomputer at the DOE that runs Linux. Nanotechnology will break past the barrier and you will begin to see Supercomputers in a testtube. Coders we need a language to code those atoms and protiens and of course it will be running the Linux Kernel and GNU Software.
How can you even compare this to a mainframe?
:) They are vastly different concepts, apples and oranges.
Clustering is a very good and very cheap and superior alternative in some cases.
In the cases where you really need a mainframe, no cluster is going to help you. Mainframes aren't even really that fast. What they are good at is having tons of I/O bandwidth, even between nodes.
If we quit comparing clusters to mainframes, then people might take clustering more seriously. They are not intended for the same classes of problems.
I have an OpenMosix cluster at home, and I work with an Origin 2000 at work. (If anyone else uses IRIX you know that you work *with* IRIX, not on it, it has a mind of it's own
I've had enough abrasive sigs. Kittens are cute and fuzzy.
From the transmeta page...
> Specifically, the RLX/Transmeta solution results in a 5x to 10x savings in power, i.e., 15 watts versus 75 watts under load, and seven watts versus 75 watts at idle.
So "lesser chips" must run at 75 watts, flat. I know Intel chips cool remarkably at idle. Remember all those Toshiba laptops frying when they're actually asked to compute? Watts is BTUs, and 75 of 'em emit a contatant amount of heat.
I hate it when they try to spin even the obvious and well know facts. If they're doing that with the black and white, what are they doing behind the Grey?
take two cabinets of these clusters, do we call them Beowulf 2.0 or Beowulf Squared? And if a cluster goes down, do we call the remaining cabinet the Root Cluster or Cluster Fucked?
I was always amused at how a unibus PDP-11 with 512K of main memory could beat the snot out of a 386 at real-world tasks. I/O is critical for so many applications...
However, don't write off clusters yet; have you looked at The AGGREGATE? The link points to Klat2 (Gort, klaatu barada nikto! Sorry) which is a very photogenic aggregate-based machine. The techniques these guys are developing may bring high I/O throughput into clustering at mainframe levels eventually.
Where did you see anything about Mainframes?
These clusters are NOT designed to take over from Mainframes, but from Supercomputers. Totally different animals.
=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Friends don't let friends enable ecmascript.
slashdot and bill mahr suck
If you want to do all your calculations in integer, go ahead, use ARM.
Wow, so it is true... Linus is a robot.
Slashdot gets worse every day... Pipedot: News for nerds, without the corporate slant
Bill Gates how nice of you to stop by and release your press release to the Slashdot community. Billy Gates "God" if you were the holy one then why are you running a email sever named "Hotmail". Mick Jagger wrote a song about you called Sympathy For The Devil. Billy Gates that was the best piece of FUD I have ever heard I laughed so hard because your so full of shit.
A blatant off-topic troll gets +5 funny, meanwhile anybody quoting Chomsky after Sept 11th gets a trip straight to -1 Flamebait. Good grief.
And before anybody calls me jealous or a whiner or a spoilsport, I WAS THE ONE WHO WROTE THIS ORIGINAL POST!!!
I WAS THE ONE WHO WROTE THIS ORIGINAL POST!!!
Was not. Quit stealing my thunder, you jealous whining spoilsporting biznitch.
Why limit yourself to the x86 instruction set when the transmeta processor just needs a new instruction set decoder to emulate pretty much ANY processor? It seems like while they'll be able to use lots of existing software out there, they could get even more performance, efficiency, or maybe just easier programming by using whatever instruction set makes sense for the project.
It's all in the pre-processing with the crusoe, x86 is just there for slideways compatibility and doesn't need to be a limiting factor. When you're using a custom computer, whether it's one or a thousand crusoe processors, wouldn't it make sense to try for some compiler efficiency based on the actual hardware instead of the 8086 legacy?
Whoa, someone should make a Beowulf clus...
er, right.
+1 Insightful, -1 Troll. What can I say, I'm an Insightful Troll.
Using this site as an example to estimate power usage, we get:
240 computer blades in Green Destiny x 6,480 hours uptime (9 months) = 1,555,200 computer hours of uptime
Assuming the only thing changed on the blade is the CPU -- and North Bridge chipset, since the Crusoe includes
a North Bridge on die and the P-III does not -- at full blast the Crusoe consumes about 1.75W of power and the
P-III + NB consumes between 4.5 - 8 W, depending on chip model. However, the 4.5W number is an approximation
from the 0.13 micron ULV P-IIIM chip running in "Battery Saving" mode, or SpeedStepped down to 300 MHz. Running
at full 700 MHz tilt, with NB, we are still talking 5.75W of power consumed.
1,555,200 * 0.0175Kw * 0.10 (dollar per KwH power cost) = $2,721.60 electricity cost/year (Crusoe)
1,555,200 * 0.0575Kw * 0.10 (dollar per KwH power cost) = $8,942.40 electricity cost/year (Intel)
A saving of approx. $6,200/year in direct electric costs.
However, the big savings comes from the heat dissipation of the units. While the newer LV/ULV P-IIIs do not require
active cooling, they still run quite a bit warmer than the Crusoe units. As a result, you don't stick a rack
full of them in a room that isn't temperature controlled. The difference in the air conditioning bill can
easily reach tens of thousands of dollars.
In business, there are two types of money/budgets. One-time grants and acquisition budgets are large chunks of
cash. Recurring expense and operations budgets are smaller. Being able to get a large chunk of cash to BUY a
cluster/supercomputer is one thing. Being able to go back year-after-year and get the funds to keep it running
is another project altogether. $15,000 - $20,000/year for electricity used in running/cooling computers is a
LOT of money to some people. This doesn't include construction or maintenance costs on a custom facility/room.
As far as reduced administration costs go, many conventional supercomputers required chilled water and other
special considerations for operation. People with experience managing things like Sun E15000s and Cray T3Es
are few and far between. They are the last of the "high priesthood" of computer administrators and cost a LOT
of money to employ.
A blade server, on the other hand, is a bunch of x86 computers running Linux -- nothing a couple of grad students
can't learn the ins-and-outs of over a term. Maintenance contracts, spare parts, etc. are also TONS cheaper for
the blade/cluster solution as opposed to high-end SGIs, Suns, Fujitsu and Cray super-computers.
Another site with a bit of good supporting information is
PC Stats.
Learning HOW to think is more important than learning WHAT to think.
I don't think there is anything special about my IP address... just tried it from two different netblocks and I can get these links just fine:
Genetic Algorithm CGI
Main Site
It's at a university - maybe they blackballed your subnet at their firewall because some loser tried cracking their systems from your site? I dunno. Maybe your browser is just busted, I'm using Mozilla 1.0 and it works fine for me.
This is seriously disturbing!
When I go to http://www.aggregate.org/KLAT2/
I get a "404 Not Found" But you get to a real web site?
What's going on here?
Does that mean that we need to evolve even bigger heads with visible veins for cooling to go beyond our current capacities? I'm thinking about Uncle Scrotor from This Island Earth, as seen in the MST3K movie, of course.
No. The veins will probably not be too visible; otherwise, a blow to the head would be more likely to draw blood. The influence of maternal instincts will demand cute babies. I predict that by the year 802701, humanity will have evolved into at least a race that looks like Precious Moments people. (I'm not entirely sure whether, as H. G. Wells predicted, there will exist another parallel lemur-like race that lives underground and eats the PM people.)
Mu-tant upgrades for all! Leaves only the fresh scent of pine!
Or, after too much mutation and crossing over, pine and human genes come together and create a little wooden boy.
Will I retire or break 10K?
could someone explain how a microprocessor is administered?
In a large cluster, the question is not whether a processor has failed, but how many have failed. Such clusters generally make it possible to swap out a failed processor while the program is running. Chips that last longer will reduce the dependency on expensive technicians to keep coming in and swapping in new boards.
Will I retire or break 10K?
Get a clue. The Crusoe consumes about 2 watts. Very nice compared to Pentium-class room heaters, yes, but I asked why they choice Crusoe over StrongARM, not Crusoe over IA-32. A 600mhz SA uses 450mW, so you can run roughly 4 of them for the same power and heat as one Crusoe.
The advantages that Crusoe has are two - first, as I mentioned originally - x86 compatibility. This is not a help for a supercomputer - you're going to be compiling everything from source anyway. The other advantage, that I forgot, is that the SA doesn't have an FPU. That, at least, is a legitimate reason to consider the Crusoe, but I'm still not sure the decision actually makes sense - the SA is a very nice chip and if programmed right it should have no problem keeping up with the Crusoe even on FP, figuring that you can use 4 times as many SAs for the same heat and power requirements.
=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Friends don't let friends enable ecmascript.
I had a beowulf supercomputer designer at my linux users group and he mentioned that alpha's were the cheapest per operation to run over any other platform. This was 2 years ago so this might be a little outdated and would be cheaper today to implement but anyway he was processor agnostic but if he did the math. The processor is only a small fraction of the total cost of the system. In this guy's example for weather forcasting modeling he had a 1 to 2 gigs of ram in each node and some expensive fiber based networking cards and switches. If you do not have at least a 1gb/sec transfer rate you have a major bottleneck. Anyway an intel based solution for his 35 mode cluster added with the networking, ram and switches averaged $2,000 a node. An alpha would average close to $3,000 a node. But he would recieve close to a %50 performance gain for using alpha's. So thats a %50 gain for a %30 price increase. Sure cooling might cost more but thats tiny compared to the amount saved by the cluster finish faster.
http://saveie6.com/
> Feng also proposed that a new technique is needed for measuring the performance of supercomputers. Instead of looking primarily at how many calculations a system can run in a given amount of time, researchers should also consider factors such as downtime, size, price and maintenance requirements, he said.
Following Feng's lead, the whole supercomputing industry has reacted to this new paradigm shift. Industry leader Cray has ceased development of its upcoming SV2 and has designed a system based on the reliable commodore 64. Explained lead scientist Joel Grey, "We managed to get a C64 computer out of the dump, and bought 1,000 surplus 'Barney' solar calculators off of ebay for $30".
The new system, dubbed the SV64, is not quite as fast as the SV2, but exceeds at new metrics: Converted to run on solar power, and having spent the last 15 years in an uncooled closet continously generating the "experiencing technical dificulties" logo for a local community access TV station, the new computer shatters existing power and reliability records. "With an expected retail price of less than $1M USD, we expect this computer to eclipse [Japanese rival] NEC's lead and become the platform that will be used to perform most of the world's weather, biological, and nuclear simulations well into the next decade", said Grey.
Wall Street analysts pointed out the the system has never needed maintence, nor suffered downtime, nor needed the services of an UNIX system administrater, and as a result, the total cost of ownership should remain low. Shares of component manufacturer Commodore rose 10 points to 10 1/64 in heavy trading today.
HIV Crosses Species Barrier... into Muppets
Imagine an imagined Beowulf cluster of one of those!
your post is interesting; because this claims that the p3m burns < 0.5W in the 300mhz "battery optimized mode". Even after throwing in the 1W your pc stats page claims for the north bridge chip, the p3m comes in under the crusoe. this claims about 1W under full load (audio/modem/etc). Interestingly enough though, the 440mx is a single component chipset, so now we have to figure in comsumption for a south bridge for the crusoe.
also, the first paper claims < 1.0W for the p3m at 533mhz, which is about equivalent to a 677mhz crusoe
anyway, its hard to judge what the total wattage each system would drain on average, and thus how much heat they would emit, but intel is much more competitive than you would lead us to believe.
280 blades in a standard 42U rack. Each blade is a P3 700 based upon the Tualatin line.
HP is continuing Compaqs blade line along with their own which will be geared toward the telco market. Also, beowulf is not really a good idea with these blades (Compaq or others) due to the need of a high speed interconnect like Myranet (sp?). Blades of these types are really only good for infrastructure and perhaps web-farms. Anything more is too much.
It's sort of hard to imagine Gordon Bell sharing a stage with Linus, at the unveiling of a Linux cluster. Isn't he the guy who absolutely loathes Unix in all its incarnations, and has been steadily trying to kill it as part of his job at Microsoft? I imagine he (and his superiors) are foaming at the mouth over the fact that Windows isn't running this cluster.
I went and read their tco estimation in their whitepaper and came across something that really made me question their conclusions.
They compare tco for 24 node clusters of different architectures of beowulfs against the bladed cluster. The biggest expense by far for the traditonal systems is sysadmin time, over half, this after they spend most of the article talking about power. They estimate sysadmin costs for each of the traditional beowulfs at $60k over a 4 year period, while the bladed cluster at $800. Where does the $800 come from? They say that they haven't had to do any maintence on their system in the 9 months its been running! That doesn't sound like a very scientific data sampling to me.
There are other bladed designs, non-transmeta based, presumably the sysadmin costs would be the same. The last chart demonstrates that sysadmin costs are what's important, and that power, space, and downtime not nearly so.
Imagine a Beowu-- agh, forget it...
Linus better watch out Gordon is up to no good. Bill Gates may be planning a sneak attack on Linux and Transmeta. My hunch is the Billy Boy told Gordon be real nice and look real hard at all those chips and schematics and bring it back to Microsoft Research we will extend whatever they have innovate it and copy it like they did with Mac Lisa and then tell the world look what we created its Microsoft Chips N Dip 1.0 running Microsoft Embedded. I would not put it past Billy to release Windows Linux Linus at the Core but all Windows on Top and LGPL the Topping. Bill Gates is ruthless and he is up to no good what was Linus thinking standing next to one of Bills Gates spies. Gordon is no friend of Linus he gets paid by Bill Gates. Microsoft Research thats where they clone other peoples shit extending it and then claim exclusive patent that they invented it. Linus did you give Gordon a Transmeta Chip well you can bet they are making one just like it back at Microsoft Research the Microsoft Way with extensions and bloat added and of course their always an extra charge called the BSA!
Imagine those imagining a Beowulf cluster of these!!!
Don't quote me on this.
Is your comma key broken?
Comment removed based on user account deletion
People who use supercomputers tend to have a clue about what their doing. They are purchased to run custom made software, like simulations, massive finite element problems, you name it, but they are not made to sit down and surf the web with or do word processing.
And finally, Microsoft knows they simply aren't competent enough to provide software for the educated elite. I can just see some astronomer loading his massive data into a big excel worksheet and trying to use the chart wizard to analyze his data. Even Visual Basic, while it is pretty nifty in some ways, is one piss-poor lame-ass excuse for a programming environment. And to think they want us to pay money for development tools too, instead of providing them as part of the OS? Grrr... getting angry... oof... eeek... turning green... ...
a version of Linux that can run on my Sony VAIO C1MV?
Imagine...
a beowulf cluster of these!
HA HA HA
BYE BYE
This appeared... how many times? In this thread alone?
Seriously though, when people make this "joke" time and time again, what do they plan to do with these "Beowulf clusters" of toasters/christmas tree web servers/coffee cans? Do they have a whole slew of PVM enabled applications just waiting to be run on some magic cluster of useless systems?
It's like all the people who want to build Mosix clusters, but fail to grasp that an application run on a Mosix cluster must not use any sort of IPC mechanisms (as Mosix can't migrate applications that use sockets/SysV message passing & shared memory)
Why do people want to build clusters? They're useless unless you're trying to model something that's CPU bound.
"Oooh, I can compile the kernel in 5 seconds!"
No you can't you twit, even with your precious cluster of 800 Pentium 75s...
he's just one of those people who keeps talking and talking until he starts to run out of air then he makes a quick pause and
sucks in some air
and then keeps talking and talking some more until he has to pause again to take another breath of air.
So, would that make it a.... Beowulf cluster?
I'm a minister!
Hmmm. It was my understanding that the 0.5W figure for the ULV P3M was in "Deep Sleep" mode. I was also assuming that when running a task, the CPU would be a full-tilt for any of the types of applications a "supercomputer" would be needed for. I see where Intel is reporting the AVERAGE power of the unit running TYPICAL OFFICE APPLICATIONS. The problem with these measurements is the CPU is 99% idle when people are typing in word -- it doesn't matter if the CPU is running at 700 MHz or 7 MHz, you aren't going to out-type it.
The ULV P3M runs a 100 MHz bus, like the 633 Crusoe but the 677 Crusoe runs a 133 MHz bus like some of the LV P3Ms.
The final problem with the P3M is the thermal diode. To control heat, once the core CPU temp reaches a certain number (100 deg F, I think -- the "maximum junction temperature"), it clocks down to reduce heat. Again, that's fine for someone typing in Word or Excel. It can clock up for the 3 seconds needed to run that macro, but for sustained high-performance computing, it will be a problem.
I'll agree that Intel is very competitive in the laptop CPU market and their LV and ULV, SpeedStep enabled chips are great in that market -- hell, I'm typing this on an IBM laptop with a SpeedStep enabled 1 GHz P3M, and it blows the doors off the Dell P3-450 I just got rid of.
However, for sustained computing where you aren't relying on user input to clock-down between, I think the fewer transistors on the Crusoe generate a hell of a lot less heat and use lots less electricity. Transmeta has some nice thermal photos on their website, but I believe they are comparing with the "old", non-SpeedStep P3M and not any of the LV/ULV stuff.
Learning HOW to think is more important than learning WHAT to think.
I don't know why you'd want to keep a machine much over 10 years, but here's a good start...
1) Replace the power supplies every 5-8 years. Overrate them by a factor of 2-3, or more. The life of an in-spec power component (caps, resistors, and such) is about 10 years. A bit more if they're run under-spec. They leak, burn, and drift with age and use.
2) When the system is new, coat all connectors with an electronics corrosion inhibitor. This oily stuff is sold to stereo freaks. (I dunno, 50W may work just as well.) Repeat the treatment whenever you replace the power supplies. When you put things back together, make sure there is zero torque being held in the sitting assembly.
3) Every year or two, break the system down and reseat all connectors.
4) Install HEPA type filtering to make sure the system remains dust free. Make sure the system has some form of *forced* air flow. Not just for the CPU, but for everything else.
5) Make sure the line power is perfect. No brownouts, no overvoltages, no spikes, ever. Even if the system survives, the components will be the worse for the ware.
6) Keep the system on, full time.
7) Don't move the system. Never move the system when it is running (even just to nudge the rack).
8) Under clock to the point you can lower logic voltages a notch. Not just for the CPU, but for everything else.
9) In computational clusters, load balance the best your problem allows. Cycles=power draw, and the components do best when the process nice steady, under-spec, power demands. Don't run problems at 100% CPU. 90% is better, so clock down, or use "cooling cycle" softare to cycle steal and "halt" the CPU for a percentage of the time. Again, not just for the CPU but for everything else.
Even with the above, there is a host of signal electronics on the board. These too drift with age, and feed degraded signals into the silicon. Not much you can do about it.
Bottom line: Run everything below spec. Keep everything cool, and at a more or less contant temp. Do not allow the smallest electrical disruption anywhere in the system. Keep power inputs as close to optimum as possible, not just in-spec.
Can you immagine a Beowulf Cluster of those?
Your math is wrong. It should be:
1,555,200 * 0.00175Kw * 0.10 (dollar per KwH power cost) = $272.16 electricity cost/year (Crusoe)
1,555,200 * 0.00575Kw * 0.10 (dollar per KwH power cost) = $894.24 electricity cost/year (Intel)
This savings is absolute dollars is much less significant when you divide by 10.
Did anyone else notice that the InfoWorld article linked every single instance of the words "Server", "Operating System", and "Processor" out to their KnowledgeLink? It's as if they feel their audience (traditionally IT crew and system admins) had suddenly forgotten these things, and so as a favor IW helped them out by providing some definitions.
...will be based on Legos.
For all intensive purposes, "whom" is no longer a word. That begs the question, "who cares"?
I know Cutler's designs from RSX-11M and VAX/VMS days. He likes clean code but he is probably less than satisfied with what happened later with NT, the amount of code that ended up running in the same space as the kernel. The original NT design was quite clean and based a lot of its ideas on Mach. Unfortunately, MS are relatively undisciplined as a company (just look at their version control problems), and eventually lots of compromises had to be built in.
Unfortunately, it isn't Linux, it is OpenVMS running on AXP. Clusters are great, but resource sharing and management become an issue. In our case, we make a lot of use of a clustered filesystem (we ensure that the data is available to multiple nodes for load sharing) and also the OpenVMS Distributed Lock Manager. Linux doesn't have this yet. Linux clusters are, as you suggest, mainly for compute bound problems at the moment.
trying to push into the kernel? Multiprocessing or Message Passing?
Coud you immagine a Beowulf Cluster of tho... Ah.. Nevermind.
their^H^H^H^H^H they're