Talk:Beowulf (computing)
From Wikipedia, the free encyclopedia
Contents |
[edit] Nearly unlimited
What is nearly unlimited? 10^10, 10^100? Non-sensical statement
[edit] Examples
I removed the two examples of @home projects because they are not Beowulf clusters. They are examples of clusters and distributed computing but not Beowulf. Lotu 21:48, 22 March 2006 (UTC)
[edit] Origin of name
"The name comes from the legend of Beowulf."
How come? The relationship between a superhuman warrior and distributed computing seems somewhat vague to me.
- I wa wondering the same myself. The closest I have found is this beowulf.org.
Billlion 15:13, 8 March 2006 (UTC)
-
- It's been six (!) years since I read Beowulf, so my memory is a big hazy. But I believe Beowulf was an incredibly strong warrior who got a Giant's sword. So he could supposedly kill any enemy, if he were 'strong' enough. A distributed computing system, if it were 'big' enough could supposedly solve any problem. (Of course, so too could an ancient, slow computer given enough time. Then again, wind blowing at the enemy would eventually blow away all of his cells). In any case, I doubt there was supposed to be a very exact correlation but if you can't see any connection between a very powerful person and a very powerful computer then you are an idiot.
[edit] What Is Funny About /. Jokes?
There are constant humorous references to Beowulf Cluster on slashdot.com. Will someone explain what the inside joke is, and what is so funny about it? User:Prophetben 8 August 2005
Nothing is "funny" about it - pathetic, perhaps, but not funny - that died a long time ago. The origin of the "joke" came about shortly after the first Beowulf clusters were announced - thereafter every time a new, more powerful CPU or computer was announced on Slashdot, somebody would say "Imagine a Beowulf cluster of those". At first such statements were valid, if somewhat vacuous - then like so many other trolls it just became Yet Another Stupid Thing To Post. N0YKG 17:19, 15 November 2005 (UTC)
[edit] More Depth Please
I was still left with a few unanswered questions - mainly, simply "Why?" - the answer I assume is performance related, but it would be nice to know more about that - e.g. optimum number of machines in the cluster, How performance increases per machine etc...
- optimum number? it depends on the task... if it is really large and easily splittable then as many computers as you can link up. if the task is small or they can't be split, then just one computer will execute best. so basically, if you have a lot of nodes in the cluster then chance is the optimum number of nodes will be reached. note also, you cannot really 'overshoot' as tasks are divided in such a way as to avoid this.
- performance doesn't increase normally per machine. the whole system's maximum performance can be much better (as a complicated sum based on processing power but limited by memory and network delays)
- As for "Why": yes, any task (computation) that requires a 'supercomputer'... eg: large prime finding, astronomic calculations, clustering (the other type, as in clustering vectors) or ray-tracing and many other graphics computations, or just to run Oblivion with milisecond load time. :D
- --Carbon Rodney
-
- Why? Consider this: you can buy a 3GHz PC for say $1000. You cannot buy a 6GHz PC, but you could pay much more than $2000 for a liquid-nitrogen cooled supercomputer. However, you can buy two 3GHz machines for $2000, network them to work together, and for a distibutable application you get the effect of 6GHz cheap. Scientific laboratories do this on a large scale, see e.g. High Performance Computing (HPC). One fun game is to maximize TeraFLOPS per dollar.
-
- And Carbon, how do you load Oblivion on your Beowulf? WINE? Could Vanguard have less lag if I loaded the client on HPC? Pete St.John 19:25, 2 February 2007 (UTC)
