Slashdot Mirror
Cringely Wants A Supercomputer in Every Garage
Nate LaCourse writes: "Real good one from Cringely this month. It's on building his own supercomputer, but with some twists." You'll probably also want to check out the KLAT2 homepage to learn more about their Flat Neighborhood Network. And since KLAT2 has been around for nearly a year (check out the poster on this page!), perhaps a 3rd generation is in the works?
Am I the only one to spot the "The Day The Earth Stood Still" reference here?
Make sure it has a red dot and says things like "Dave, what are you doing Dave?" Can't wait for mine!
Satanists get good grades too...suspiciously good grades
A story that beowulf cluster posts will be relevant to!
Remember, there were no nuclear weapons before women were allowed to vote.
This is a very interesting concept that he is putting forth, but at the same time, how many geeks out there are going to really make use of such a clustering farm? Not everyone I know does video compression projects, and it would seem kinda prohibitive for a black-hat to set one up to break encryption codes. Could someone please tell this naive soul what useful everyday application all these CPU cycles could be used for? (if you say SETI@Home, I am going to bitch-slap you)
Secondly, UWB seems to be the holy grail of wireless networking, yes, however is this something that the agencies of the world are going to let out of the bag so easily as he says, I can think of the CIA and the NSA having a few choice words about such "undetectable signals" being used by commonfolk after September 11th...
Just my two cents
ZDNet has an article of HP building a supercomputer like this as well, called the "I-Cluster." It has 225 networked computers running Linux Mandrake (so changes could be easily made) on 733 MHZ out of the box PCs. The only catch is that is is slightly more expensive- $210,000 (minus network cabling). On the other hand, they plan to release the open source tools they made as well, so that people can repeat this.
"I have not failed. I've simply found 10,000 ways that won't work." --Thomas Edison
And it shall be called. . . Earth!
You want the truthiness? You can't handle the truthiness!
I always think that it isn't worth to waste the valuable garage space on my second-hand japanese car, which worth no more than $1,000.
Now it is used to place a $41,000 supercomputer! Ph43r m3!!
but then, I wouldn't allow anyone driving a car into my garage(WATCH THAT NETWORK CABLES ON THE GROUND!), so should I build another garage for my real cars?....
Those are some interesting ideas.
Now how about organizing them before publishing them? Call me pre-postmodern (and I'm still in my twenties), but I tend to learn more from a coherently-organized message than from a random jumble of statistics and facts. Cringely jumps from a detailed description of the KLAT2 and its innovative networking technology to a brief description of UWB. And then it's over.
Maybe I'm missing something.
"First you gotta do the truffle shuffle."
Well, not to be one of those stick in the mud 'Read the %($#ING article' type people, KLAT2 is a reference to The Day The Earth Stood Still. Had you looked at the articles in question (particularly, the KLAT2 page) you would have discovered that indeed, they were intending the reference. Heck, go check it out - the poster they made up for it is worth the look! :-)
Davis Ray Sickmon, Jr - looking for something to read? Check out my three free novels at MidnightRyder.org
What good is a supercomputer in your garage if you do not use it to maximize garage-holder value? If you provide supercomputer habitat for the progeny and supercomputer embodiment of the JavaScript AI Mind, which has also been coded in Forth as Mind.Forth Robot AI, then your home-sweet-home garage will be a major waystation on the road to the Technological Singularity.
Just as the Shroedinger Equations for atomic bombs and such were developed seventy-five years ago when Erwin Schroedinger spent his 1926 Christmas vacation holed up in the Swiss Alps and working out a few mathematical formulas that shook the world, nowadays over the 2001 Yuletide there have been the first stirrings of True AI in the JavaScript AI Mind, which any garage tinkerer may adapt for either 'pert near all-powerful supercomputer AI or a killer-app if not killer robot.
Following in the footsteps of the giants who created Visual Basic Mind.VB and Java-based Mind.JAVA, be the first on your block to create the supercomputer-based Garage-Mind.
Maybe they could set things up so that ALL his articles hit the main page as soon as he posts them.
If this were the case he could put a "discuss this article" link on his page and simply link to /.
Lasers Controlled Games!
Sheesh...!
You're using her as bait, Master!
I'd really like a cheap farm for video compression. I tied up a G4 433 for six hours last week compressing a 20 minute movie using Sorenson 3. Fortunately was using OSX and so the machine remained relatively responsive, but still, six hours pinned just for 20 minutes. (Of course it did take a 4GB movie down to 150 MB)
:-) I imagine Apple won't be porting iDVD to Linux anytime soon, and the stuff the studios use is either custom or very expensive.
So now that we have a cheap supercomputer, all we need is cheap software.
Btw, a little nitpick, the TX would refer to 4 pair (all 8 conductors), 10base-T uses 2 pair, 10base-TX and 100base-TX use all 4 pairs.
Still though, after having to wallow through Cringley's painful lack of comprehension of basic technical knowledge, reading the ArsTechnica piece again was quite refreshing.
Is your company running tools written by ma
Through Google I found the UWBWG, and there's lots of detailed papers at Aetherwire. Interesting reading.
What do you think of MusicCity now?
Cringely is completely missing the point. KLAT2 uses multiple routes and switches, not channel bonding. And what the project contributes is not the basic idea of using multiple network interfaces (which is decades old), but a specific approach: using genetic algorithms to optimize the network topology. More traditionally, such clusters have used manually designed topologies with known performance bounds.
Back about... 15 years ago? 20? Something like that.
I was in the office of a research company and the owner showed me their shiny new minicomputer. I can't remember what kind it was, unfortunately.
He said something then that struck me as very insightful and I've not forgotten it to this day.
"You know, minicomputers are looking more like micros every day."
If you're a zombie and you know it, bite your friend!
Speaking as someone who, yes, has actually worked with the big iron...
Why bother. Remember, Moore's Law is still in effect. Recently, we've hit the point in the curve where supercomputers are no longer needed, nor cost-effective. That is, the time it takes for the industry to deliver a far superior product has eclipsed the average lifespan of your typical supercomputer.
We're living in an age where a single graphing calculator you can buy at Walgreens has more horsepower under the hood than what got us to the moon 30 years ago. Your $2700 PC will be worth $150 within 3 years.
Having a supercomputer in every garage makes about as much sense as taking a rocket fuel-powered dragster to the supermarket for a gallon of milk.
Cheers,
Bowie J. Poag
I'd rather have a superMODEL in every garage.
7 November 2006: The day Americans realized corruption and incompetence weren't addressing 11 September 2001
The costs of a clustering setup go well beyond the initial hardware. At the level that Cringely is building (with only 6 machines), it may not be a huge problem, but running KLAT2 will cost you some dough just for the power.
A couple years ago I made a dumb mistake and bought a saltwater reef tank without realizing that it would end up costing me $150/mo. in electricity bills (it ain't cheap running 4000+ watts in lights and pumps 18 hours a day). I'm sure running 66 machines 24 hours a day ain't cheap either.
Also, if everybody had a supercomputer in their garage, they would no longer be so "super."
It used to be that saying "10 base X", where X was a number, implied that the medium was coax cable where the maximum length of the network was given by X. However, usually X was given in hundreds of feet, such as "10 base 2" or "10 base 5". This could yet be an error ... haven't seen coax used in a real network for years.
Toronto-area transit rider? Rate your ride.
Dr. Dietz used to teach at Purdue, and I had the good fortune to take a compiler course taught by him. On the first day, when introducing himself, he came to the part where he was describing how to get into contact with him. When giving out his phone number (at Purdue, on-campus numbers were 5 digits long) he mentioed that his phone number was "GEEKS". He added, "No, I didn't ask for GEEKS, but when I figured it out, I thought it was pretty cool."
Needless to say, it was a pretty cool course.
25 dual p650s in my home office ... when I crank it all the way up I come in at somewhere in the 50-100 range on the dnet rc64 dailies. Sadly the original reason I built it has evaporated and with the current cost of CA power I just have a fraction running
<disclaimer>I know little about 'big iron'</disclaimer>
But isn't the point of these kind of projects to derive more computing power in a generic form, something useful to many situations?
Sure, my Athlon isn't too slow at the piddly little hobbyist 3d rendering stuff I play with, but what if I suddenly get grandiose dreams of 3D worlds, wouldn't it be nice if I could divert the down payment for a house and move myself a year or two farther along Moore's timeline?
I can think of some small business applications where a nice quick video compression would be nice, especially if the hardware and software were all generic enough to buy off the shelf without a serious outlay of cash. Granted, there are very nice and very fast hardware codecs but then what if that same small business wanted to render some 3D along with that video stream? Or I'm working for them and get permission to render my VR opus overnight?
What about applications that could be enabled by cheap and standardized GFLOPs? If you can't think of any you're not thinking hard enough.
Bleh!
Clusters are great for embarassingly parallel applications (ie ones that have threads which don't communicate with each other much. This includes things like SETI@home and batch rendering of images. What they don't compare on is applications that communicate a lot like nuclear physics simulations. This is not to say that that will never change in the future, but for the time being it's still true.
Last, and certainly not least, real supercomputers have memory bandwidth that can match the speed of the processor. A Cray or an SGI Origin has an absolutely massive amount of bandwith from the processor to local memory compared to a PC. That allwos a traditional supercomputer to actually *achieve* the fantastic peak performance numbers. On many applications, the working sets are huge and don't fit in cache so you end up relying on memory being fast. On a PC, it's not and I've heard from sources I consider reliable (though I have no actual numbers to back this up so it may be rumor only) that one large cluster site sees around 10% or less of peak on a cluster for a nuclear physics simulation, whereas, on a vector Cray, you can hit ~80% of peak. This means that the cluster has to be 8 times more powerful and when you start multiplying the costs by 8, they start looking like the same price as a real supercomputer.
So my point is that building a real supercomputer does not mean grabbing a bunch of off-the-shelf components, slapping them together with a decent network and running Beowulf (or a similar product).
Go Badgers! -- #include "std/disclaimer.h"
The FCC is being very cautious about mass-market UWB products. Since these things blither over a gigahertz or so of spectrum, they overlap with other services. At low power, a few of these things are probably OK, but in bulk, there could be trouble. The concern is that mass deployment could wipe out other services in congested areas.
...a single-CPU version of this!
One: Because we can.
Two: Ever seen the stuff they run on supercomputers today? Simulating a supernova for 1 nano second can take a month of CPU time on some of the world's fastest supercomputers. Oh, its still very nessesary. If the past is any indication of the future, we will always need blazing fast machines to push the limits in the scientific world.
I assume you mean big iron as in mainframe, which is NOT a supercomputer by any means. Mainframes do the work that runs this world, supercomputers help us discover what we'll do in tommorow's world. They are very different worlds.
--Nuintari
slashdot : where an opinion can be wrong.
--Nuintari
slashdot : where an opinion can be wrong.
Gb Uplink ports wouldn't really help - the traffic pattern inside KLAT2 is flat, where all nodes are equal. Not like a LAN where it helps to not have a bottleneck at the switch interconnects.
~~~~~ BigLig2? You mean there's another one of me?
"(the operating system) will be QNX, a real time OS that supports massive parallelism and has very low overhead. QNX is fast! QNX is also Posix compliant, so there is lots of software that almost works under it."
If you're looking for software that almost works, I know of an OS that might fit your needs. You're not going to hook this thing up to the Internet, though, are you?
Yes, and the 'T' in 10BaseT stands for 'T'wisted Pair.
How fast are those cards actually?
I know D-Link's PCMCIA 100BaseTX cards are 16-bit, so while they will signal at 100MB/sec, their throughput is not any more than (as far as I can see) than you would get from an old desktop NE2000 adapter. Low end network hardware frequently pulls this kind of stunt -- repackage old technology so that it will look like it should perform better than it actually can.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
By definition only the fastest devices are supercomputers. These days that is about a teraflop. Thta includes the US DOE ASCI series and the announced installation of the Blue Storm and Blue Gene IBM computers. Ten gigaflop computers a dime a dozen and a hundred gigaflops not so rare.
Yes, these systems are not sometimes the best for handling vectorizable jobs, but they are so inexpensive compared to the old specialized hardware that it is easier to waste cycles than build special hardware.
As to memory bandwidth. Modern CPU caches make the question nearly moot.
If all of this were not true, then people wouldn't be building clusters and the majority of the top500 list wouldn't be dominated by clusters. Instead there are 3 traditional architecture machines in the top 20. This is the reason that Cray (etal) no longer dominates the marketplace... commodity systems have overtaken nearly all of the specialized hardware world.
-- Multics
It wouldn't help so long as the CPUs couldn't utilize the gigabit bandwidth. Swap out the 100 megabit lan cards for the cheapo gigabit ones for slightly more money- I think you'll find that this cluster's still starved for bandwidth.
I am not merely a "consumer" or a "taxpayer". I am a Citizen of the State of Texas
>But what do you expect from someone who practices identity theft [wired.com]?
How exactly do you steal the identity of someone who never existed? The man we know as Robert X. Cringly was the Infoworld Cringly for 8 years! I'd say he pretty much defined who that Cringly was (or is today, I don't read Infoworld.) Saying he practices identity theft would be a valid argument if the Infoworld Cringly was someone else and he had just appropriated the name for use on PBS, but he didn't. He built up the Infoworld Cringly and so I believe he has a right to go on with the persona he's used for all this time.
man RTFM
No manual entry for RTFM.
I have a TiBook. Apple *could* do away with the screen, keyboard, and speakers, and replace the CD-ROM slot with a ram-bay.
:)
Not only could you hook them together using gigabit ethernet, you could take advantage of the firewire port as well, perhaps chaining them together with some sort of SAN, though you are still limited by the ~50MBps, though perhaps that's not useless, I don't know.
Still, with the ram bay you could up the memory from 1GB to something crazy, like 16GB. The battery is useful as a backup-emergency device, allowing the slab to run for about 4 hours in case of emergency (woo!).
You could even concievably netboot the thing, since OS X allows for that, right? Minimize the hard drive or get rid of it altogether... you could seriously make a slab about the size of 1/2" by 8" by 8" I suspect
GPL Deconstructed
1) heat in garage in winter
2) Top 10 in Seti@home
3) Porno-ize you favorite anime (Final Fantasy anyone?)
4) Why are you reading this? I thought you were doing #3
make Linux, not Microsoft. sin(beast) = -0.809016994374947424102293417182819
All I know is that it's a G4 at somewhere in that range. Maybe it's 466? It's gray and white, and I bought one of those nifty 17" flat panels for it.
It was idle and had been running OS 9. I swiped it and installed 10 on it because I needed to produce a video in short order and didn't want to fuck around with installing a firewire card and Adobe Premier in my NT workstation.
Really iMovie seems to be a cool program. It was no problem to import the video, cut it up, resequence it, add transitions, sound, etc. The only problem I had was that the output for full-screen high quality was 4+ GB, and compressing it so that it would look good took me several tries of multiple-hour conversions.
Thanks! I'll try it out when I get back to work.
This movie project was my first experience with OSX, and the first real time I've spent with a Mac since I gave away my 7100 a few years ago... With so much control over the OS and a system that didn't crash on me once I think I'll be spending more time with it.
I guess your ideas are kind of on the right track, but you should probably familiarize yourself with modern system architecture trends.
;-)
Crossbar-style system interconnects are not new ideas. I'm not an authority on the subject, but I know that the Cray Y-MP had a 32-port switch architecture that provided about 1.3 GB per second of memory bandwidth per processor (hope I'm remembering these numbers right!)
The DEC VAX 9000 series had a 1 GB/second CPU-to-memory pathway that utilized a crossbar switch, also.
Both of these systems were in wide use around 1990, give or take a few years. And, of course, the ideas go back much further than that. I used to have a copy of a paper by Wulf in Communications of the ACM dated 1974 that described a switch-based multiprocessor system. Can't find it right now, alas.
Things have come a long way. From 1 GB/sec aggregate in 1990 to 22 GB/sec aggregate in 1998 (the Cray SV1) to 40 GB/sec aggregate in 2001 (the SV1ex). The SV1ex provides each processor with 6.4 GB/sec of bandwidth into and out of main memory.
Increasing the speed of the RAM isn't the issue-- the SV1ex uses commodity SDRAM. The issue is building sufficiently large parallel paths for the memory controllers to execute very large parallel fetches into a vector cache.
So I guess you could say that you're headed in the right direction, but you've got a long way to go.
Uh, Cringely, wouldn't creating the thing and then using it as the subject of an article for the company that employs you count as a commercial purpose?
You don't really expect QNX to bitch about a little free advertising do you?..