Slashdot Mirror
Linux Clustering
An anonymous reader writes "Beowulf clustering turns 10 years old, and, in this interview, creator Donald Becker talks about how Beowulf can handle high-end computing on a par with supercomputers."
← Back to Stories (view on slashdot.org)
... the amount of replies that will start with the same subject header as mine and not be funny at all?
I sure can!
Good quote, too many chars. Seriously, the slashdot 120 char limit sucks!
i dont mean to sound like a troll or anything, but is this really news. over the last year or so, (nearly) all of the articles on /. about fast computers have been clusters.
awww, fuggit
I'm picturing the ten candles on the Wolf-cake in close proximity with frosting interconnects and one big flame in the middle.
What if Digg added local news and a Slashdot inspired comment karma system? ---
http://houndwire.com
Happy Anniversary to the most over-used joke on Slashdot. I'll be wearing my tin-foil hat all day to commemorate it. (The 10th anniversary is the aluminum/tin anniversary)
All this sounds good and Interesting, and Becker did a tremendous ammount of development in this field. But I was just wondering, what about supercomputers like BlueGene/L which have very fast interconnects. Many supercomputers/distributed systems run MPI based programmes and such programmes need a high interprocess commmunication does anyone one know how good these are in a Bewoulf Cluster? thanks a3217055 They said that of all the kings upon the earth he was the man most gracious and fair-minded, kindest to his people and keenest to win fame. :-The Geats' tribute to Beowulf after his death.
...doesn't it to you? I mean how long have you been sick of the "imagine a beowulf cluster of those" comments? Doesn't seem like only 10 years would make me that sick of it.
Weaselmancer
rediculous.
Could it be that Beowulf clusters, however cost-effective and powerful they have become, are passé now that most universities and research institutions have some sort of COTS-based high-performance computing solutions? Not that Beowulf isn't cool - it is - it just doesn't seem as cool as it used to.
...can be simple. The more complex a problem gets, the more likely you need one supercomputer as opposed to a cluster. It's not elitism, it's just that the problem will probably require a lot of communication between processors.
Any kind of networking solution between computers will never be as fast as a hard-wired bus can be. If a lot of communication between nodes is required, you will spend more time waiting than computing, which shoots efficiency to hell.
imagine a lone computer sitting by itself not connected to anything...
I'll turn into a supernova and burn up everything. Well I'll turn into a black little hole and you'll turn into string.
If you could imagine a...ok, well maybe after ten years, we all could. The horse has been so beaten and tenderized that even takko vell wants a piece of the action.
I've never seen a beowulf cluster personally. I've never run anything on one. However I do know that it made "supercomputing" more affordable. That in itself is a feat -- and a primary goal of most Open Source software. A proverbial "Hats off" to the open source hackeres out there. Thanks...and keep hacking.
Now if I can gather enough old 486 machines, will I be able to run Doom III? With WINE? At least it'll keep me warm during the winter months.
In Soviet Rus-- oh damn. I pulled out the wrong dead horse.
Quote from the article: *snip!*
I was wondering if it is possible to make some sort of cluster out of old computers I have lying around? Nothing spectacular, just hooking up 3-4 old P2's to make a game server or something of the sort. Is there software out there to do this?
Has anyone had any experience with this?
Just a thought...
what is the most over-used joke on slashdot? options:
1. step 1: [topic], step 2: ?, step 3: profit!
2. natalie portman/hot grits
3. in soviet russia, [inversion of topic] you!
4. cowboy neal [action]
5. beowulf cluster
6. goatse guy
7. [technology/entity] is dying!
just curious...
ed
Clusters are the slowest computers available...
If your metric is moving around data, as opposed to how many no-ops you can do a second while waiting for your data to get there.
paintball
processing...
To be considered a "supercomputer," it also needs enough CONTIGUOUS MEMORY SPACE to hold the massive amounts of data associated with true "supercomputing." So far, no cluster has met that requirement.
That is what I want to know!
In all seriousness though, what is the ratio of cluster to big iron in supercomputing nowadays? I know a clusters can scale out to a lot of FLOPS, but what is the highest FLOPS processor available?
-Randy
Imagine there's no cluster,
It's easy if you try,
No adapter below us,
Above us only loopback,
Imagine all the computers
computing for themselves...
Imagine there's no internet,
It isnt hard to do,
Nothing to download or upload for,
No porn too,
Imagine all the computers
computing pi in peace...
Imagine no tokens,
I wonder if you can,
No need for ethernet or tcpip,
A brotherhood of computer,
Imagine all the computers
Sharing nothing at all...
You may say I'm a dreamer,
but I'm the only one,
I hope some day you'll leave us,
And the computers will computer alone.
Mid-Eastern Pennsylvania Gaming Convention
And GridEngine is free and opensource:
http://gridengine.sunsource.net/
Donald Becker also has done a large amount of work on Linux Network drivers. Grep through linux/drivers/networking and you'll find he's done work on Intel NICs, Realtek 8139s, even the ne2000 (I think he said he puked a few time while working on that one). Thanks for all your hard work Donald!
At the end of the article, the comment is made that one reason for setting up a cluster is ease of management (for updates, applications, etc.). Can anyone with experience comment on whether this is true or not, with the way clustering exists today? I have no experience at all with cluster, and I'm wondering if this is something I should look into to ease administrative burdens?
Ok
The Tao of math: The numbers you can count are not the real numbers.
Beowulf isn't the only game in town folks. A much easier to maintain and balance cluster can be built using openMosix. openMosix is a single-image-cluster extension for Linux.
Full-Featured GPL Web Hosting Control Panel
There are certain classes of problems that clusters don't map to well. Applications with a very high cost of inter-processor comminucation or that demand a huge piece of contiguous memory are probably always going to be outside the realm of clusters.
However, problems that are embaressingly parallel can be handled by a cluster very adequately for a fraction of the cost of a traditional supercomputer. I don't know that you can ignore this class of problems and say that clusters aren't "true 'supercomputing'".
If you maintain a group of networked but otherwise independent computers for example a student lab or office farm, consider deploying something like PVM or MPI. It's a great way to get some use out of those idle cycles.
PVM at least scales incredibly well: 25 machines rendering a povray scene take just a fraction over 1/25 the time taken to render it on one machine. I haven't tested MPI yet.
"Nine times out of ten, starting a fire is not the best way to solve the problem." - my wife
- Can you imagine a Beowulf cluster of these?
- How long until the RIAA sues them into oblivion once they find out how may MP3's you can put on one?
- "Can you put Linux on it?" or "Yes, but will it run Linux?"
- "Yeah, but does it run Doom3?" or "And it still won't run Doom3."
- Any comment regarding "Duke Nukem Forever" taking literally 'forever' and being termed 'vaporware'.
- I am not buying one until they support ".ogg".
- I for one welcome our new (insert name of company mentioned in post or story) overlords.
- "George Lucas raped my childhood" or "Greedo shoots first" comments on any story incorporating the Star Wars franchise.
- A comment including these 3 components in any order: Natalie Portman, naked and petrified, hot grits, one's pants.
- Microsoft = Evil, MPAA = Evil, RIAA = Evil; with anything else incorporated to try and fit those equations into the topic at hand
- Some type of reference to the size of one's ProN collection, the amount of ProN that can be stored on the gadget or technology in question, or the ProN industry itself being the first to make good sue of the new technology or gadget in question (ergo: the ProN industry drives technology)
- The posted cliché being self-described as an "obligatory" post in the heading area if that particular cliché had not been addressed yet by previous slashdotters. (e.g. "obligatory Beowulf cluster comment")
- Post revealing the fact that the story's homepage had been slashdotted already, culminating towards another post later on with the homepage story itself being copied & pasted verbatim (often with a subsequent post purporting that this is karma whoring, even though the poster admits it is indeed helpful anyways.)
- Remark on the size of some new storage advancement about how many LOC's (Library of Congresses) can fit on it, or any other remark noting how this can be an actual valid unit of data storage measurement.
- A variation of the Zero Wing video game intro dialogue regarding it's broken English translation: "Someone set up us the base....we have every ZIG, make your time".....blah, blah, blah.
- Very soon lists such as this will be clichés as well.
- Similarly noted and additional clichés may be found here: http://en.wikipedia.org/wiki/Slashdot_subculture
>>>>>> Chewie, take the professor in the back and plug him into the hyperdrive.
Sad to see this little knowledge about parallel computing on slashdot: blatantly wrong information marked as informative. +5 no less.
Let's address this first: there are two common memory architectures, distributed memory (a cluster) and shared memory (a 'traditional' supercomputer). Each can emulate the other. Saying a cluster doesn't have enough memory, presumably at each node, is really saying: "I don't really understand message passing."
This would be more important if datasets were actually large. Unfortunatly for your argument they aren't. A handfull of nodes and they'll hold the whole simulation easily in memory (albeit it'd take years to run because there's so few CPUs at work.)
How would I know? Well, I work with the Center for Simulation of Advanced Rockets aka CSAR at UIUC, one of five DoE ACSI sites in the country. I manage their supercomputer, which is getting upgraded from 200 P3-class dual proc PCs to 640 dual proc Xserve G5s. Before that I was a grad student working with them, albeit not on the CSAR simulation but instead on a related grant, the CPSD.
Now, there are computing problems which clusters aren't good at (or at least that's the traditional claim. My master's thesis and advisor would seem to dispute that this is actually the case.) However, most problems as the interview says, run just fine on clusters. Physical simulations (which covers CSAR's rockets to the national labs nuclear weapon research to hurricane/weather simulation, all the way down to protein folding and atomic and sub-atomic scale crystal formation simulation) need to know about what's in the area you're working on, and what's in nearby areas.
Occasionally you'll find an oddball like galactic simulation (or molecular dynamics) that needs to compute gravity across the whole universe. Fortunatly we have multigrid methods and a friendly gravity equation to solve this problem: get real data from those near you. Average those far from you and use that instead.
Then of course there's the idea that even "traditional" supercomputer problems that don't run well on clusters can be run efficiently on clusters IF you move beyond 1 process per CPU. Load up 10, 20, 100, 1000 little workers on a processor. Get fast context switching between them (not OS level!). Use message passing rather than shared memory (locking, ick!) to communicate. One worker blocked waiting for network data? Process the next one! If you've tuned things right you'll find you always have work to do.
Sounds crazy? Supercomputing '02 didn't think so: http://charm.cs.uiuc.edu/research/moldyn/
Slashdot Patriotism: We Support our Dupes!
ClusterKnoppix is pretty cool. Its got all the auto detect hardware features of regular knoppix, and also aautomatically adds itself to the cluster.
In Soviet Russia, Beowulf Cluster imagines you!
Why use Beowulf when you have openMosix? openMosix is all transparent to your application. You dont have to worry about remote execution, openmosix migrates your process automatically to the best node. Like I said.. it's all transparent and requires no additional programing in your application.
Though John Lennon was a Beatle, Imagine is not a Beatles song.
Rocks provides an easy way to build a Beowulf cluster. See http://www.rocksclusters.org/.
. shtml?tid=29&tid=94 for one account.
You can build a working cluster, starting with the hardware and installation CD-ROMs, in minutes; see http://servers.linux.com/servers/04/08/27/1943227
Disclaimer: I work with the folks who created Rocks.
I noticed that this article by the PenguinComputing CTO appears to answer the article by the Cray CTO and contradicts it. All I want to know is this: how much did PenguinComputing and Cray spend on advertising banners on SearchEnterpriseLinux to have these articles made? Let's let vapor settle before it gets to our heads.
-Those who would give up essential liberty to purchase temporary safety deserve neither. -Ben Franklin