GeneralThis forum is for non-technical general discussion which can include both Linux and non-Linux topics. Have fun!
Notices
Welcome to LinuxQuestions.org, a friendly and active Linux Community.
You are currently viewing LQ as a guest. By joining our community you will have the ability to post topics, receive our newsletter, use the advanced search, subscribe to threads and access many other special features. Registration is quick, simple and absolutely free. Join our community today!
Note that registered members see fewer ads, and ContentLink is completely disabled once you log in.
Blue Gene/L compute nodes use a minimal operating system supporting a single user program. Only a subset of POSIX calls are supported, and only one process may be run at a time. Programmers need to implement green threads in order to simulate local concurrency.
My amiga 500+ was slightly better, it could run more than one process at a time
i was having a read about IBM's Blue Gene stuff the other week. They do sound very interesting, but don't be mistaken by the quote you left sycamorex.
Quote:
From wikipedia:
Quote:
Blue Gene/L compute nodes use a minimal operating system supporting a single user program. Only a subset of POSIX calls are supported, and only one process may be run at a time. Programmers need to implement green threads in order to simulate local concurrency.
My amiga 500+ was slightly better, it could run more than one process at a time
The machine as a whole does not run one process. Each node within the machine runs a process. Without going and looking again, they have ALOT of nodes. Further more it can be sub-divided in a number of ways. So what is one whole machine can in fact look like several virtual machines, with each virtual machine having multiple nodes in it. Most of it went straight over the top of my head, but i understood roughly how it works.
In effect its like having a room with thousands of Amiga 500+'s all linked up on a network with each one running a single process but still working together. Even so it is still way beyond the power of probably all the surviving Amiga 500+'s that are still left in this world
i was having a read about IBM's Blue Gene stuff the other week. They do sound very interesting, but don't be mistaken by the quote you left sycamorex
just kidding
Quote:
It was recently upgraded and now runs at a speed of 478.2 teraflops (trillions of calculations per second), using 212,992 processors.
On a serious note, what puzzles me is when you have over 200 000 processors and they are physically distributed over a relatively
large area, aren't benefits of having multiple processors slightly reduced by the physical distance between them (the time for electrical impulses to travel). How would one 'big' processor deal with it? I guess there would be more problems with cooling and a much higher cost of such a processor, but in terms of efficiency would it be better?
In some ways there are disadvantages over the distance between each processor. But then these machines aren't built for the kind of applications we use everyday (by we, i mean us mere mortals who just use computers to run "simple" programs to amuse ourselves)
To give a very very simple example, imagine doing you 2 times tables, 1x2=2, 2x2=4 etc. now if you had 1 processor doing this and 1 sum takes 1 second and you wanted to go to 10x2 it would take 10 seconds plus say 2 second overhead to sort the results, thats 12 seconds altogether. Now how long would it take 10 processors? 1 second. Looking at the example then yes, but then again you allow some overhead for bringing all of these sums together into a list you could say add 2 extra seconds to that, so it would say take 3 seconds to complete. Now compare that to 12 seconds for one processor. What do you see?
That was a very very simple example! But you'll have to forgive me about that, as i'm no expert, but i do understand roughly how it goes together.
If you think along these lines and instead of doing that simple sum it is doing something like the folding project (probably a bad example) you can see how time is being saved. After all that is mainly what having multiple processors is all about, time.
Amongst other things having a set up like the Blue Gene's. If a node fails it can be isolated from the rest until a suitable time to fix it.
To anyone else reading this, who does actually know what there talking about. Please tell me if i'm just talking rubbish or not?
To give a very very simple example, imagine doing you 2 times tables, 1x2=2, 2x2=4 etc. now if you had 1 processor doing this and 1 sum takes 1 second and you wanted to go to 10x2 it would take 10 seconds plus say 2 second overhead to sort the results, thats 12 seconds altogether. Now how long would it take 10 processors? 1 second. Looking at the example then yes, but then again you allow some overhead for bringing all of these sums together into a list you could say add 2 extra seconds to that, so it would say take 3 seconds to complete. Now compare that to 12 seconds for one processor. What do you see?
I think that's a very good example. You wouldn't run things like, say a word processing application on something like that.
Some of the fastest supercomputers in the world run beowulf, a Linux version for high performance computing. Linux is everywhere these days. A few years ago there was a Linux version running on a wrist watch, then there is Beowulf and of course all the desktop things in between.
LinuxQuestions.org is looking for people interested in writing
Editorials, Articles, Reviews, and more. If you'd like to contribute
content, let us know.
