View Full Version : Beowulf clusters
Well, I gotta set one up for a school project as proof of concept.
I've been to a few sites got tutorials, beowulf underground and the beowulf project. Anyone got an advice, tid bits or links of their own that they recomment.
Currently I have weak hardware, p133, 32 megs ram and about 1.2 gigs of space per.
Again this is just, can you set it up, type deal.
Any help is appreciated.
I am curious, How is the project going?
Well I have two boxes up and running on Linux, one is 5.x (all I had at the time and I was curious about hardware support straight from the CD) and the other is 7.2, both Redhat. Yeah, I know 7.3 is out, but I'm more concerned with getting stuff up and running.
I haven't done much else, I gotta see if the two machines can see each other, after that, I suppose I should do the NFS stuff and then see what I have left. I'm also going to have to figure out what to run to show that my beowulf cluster isn't a big ol' waste of power.
You have a good setup. Please keep us informed with the progress :wink:
BTW, I just bought the RedHat 7.3 (2 CDs) in the newspaper stand for US$ 4.50 and will try to install it in my computer (multiboot with WinXP). I will try to learn about linux slowly.
My friend and I did a lot of clustering back about 2 years ago when I was doing a lot of stochastic raytracing. Unfortunately I didn't really deal too much with actually setting up the cluster itself (they were all his machines, and in another state).
If I remember correctly, with beowulf you'll need to use software that is written specifically to use PVM for the load balancing and distribution to take place. This is probably ok for stuff like povray and some of the other distributed apps that have compiled versions to be used with beowulf. Otoh, a better proof-of-concept show may be to use mosix. Mosix is clustering software for linux that basically allows processes to be distributed between systems (and cpus) in the same way that you'd expect SMP to work on a multiprocessor system. Basically you tweak the resource usage values (how important bandwidth vs processor vs memory is to you) and it automatically farms out processes to whatever machine(s) seem to be the most worthy. What's probably more impressive, is that it works with reguluar multithreaded applications (I think this includes both kernel and userspace, but not sure).
Mosix is what we used on the cluster for distributed rendering, and it worked fairly well once we got the parameters set right. It can get kind of bad when it doesn't know what do to about resources (say it runs out of ram on a node) and starts trying to move the process around to different computers, hosing the network. Otoh, overall I was pretty amazed at how well it seemed to work. Pretty neat that it could have seti running on all the cluster nodes, we could tell it to migrate the process off to another node, shut down the system, boot to windows, play halflife for a while, reboot the system to linux, rejoin the cluster, and migrate the thread back to that node as if nothing ever happened.
First thanks for the replies. Pascal, I'll be sure to keep posting.
First off, thanks for sharing. Second, that sounds HELLA kewl. I'm gonng have to start messing with it immediately.
What Linux kernel version were you using, how ugly would it be if I installed mosix on a 2.4.7 (Redhat 7.2) when they say they support 2.4.18?
Could you download the ISO images of RedHat 7.3 and burn some CDs?
Maybe it will help you.
Heh, I'm doing that as I type this message out. Though, it's going to take a while. Only got a T1 with an ass load of users on it. =(
This was back about 2 years ago when we were doing this regularly. I believe we were actually using one of the 2.2 variants at the time.
I'd imagine that 2.4.6 would probably work out ok, so long as there aren't any specific things about kernel 2.4.18 that it wants. You could always email one of the devs. :) Otherwise, is there any big reason you need to keep 2.4.6? You may get a nice boost in performance with the new VM that came around in 2.4.12ish.
Heh, I"m not up to date on the Linux developments of late. Anything else that's neat besides the new VM?
Hrm, probably, but I haven't followed it very closely either. The VM thing was rather big though, as a lot of people were giving Linus flak for completely overhauling the VM in a point release stable kernel. :)
2.6 looks to have some really neat stuff in it though, like (again) a new VM, o(n) scheduler, serial-ata support, LVM, EVMS, etc etc.
Btw, how's the project going? Haven't heard anything in a bit.. :)
Well I got them (computers) seeing each other. YAY. So far I've got so good. I was messing around with NFS, because I was following COCOA (I believe that's how it spelled) tutorial, which used RH's kick start to install the clients from the Server.
I think I might just use the usually MPI libs and make myself a quick program. Or get something of the net. Suggestions, PLEASE!
Other than that, it's due on this Friday. ;)
I hear Linux 2.6 is suppose to have some really spanky multithreading and multiprocessing performance improvements. Finally, it looks like it "might" be competition for BSD. ;)
All right, so I got two nodes setup. Installed an MPI library distro, called MPICH. Unforunately, my RSH won't work! Gah!
Wll damn it, I forgot to include the rsh and rlogin into the tty1 file so it wasn't working. ;) There were a few other kinks, they finally got resolved.
Then it was onto NFS, stupid thing wouldn't mount it via the DNS/aliases so after hardcoding the IPs it worked.
So, the nodes were up and running and I used CPI (c program to approximate a 32bit pi). It ran about 10,000,000 iterations in about 34 seconds on the two nodes while doing about 68 seconds on a single node (server node to be exact). I couldn't run much else, since I didn't have the graphics stuff installed.
Congratulations Saem :wink:
I installed the RedHat but the Gurb (redhat but) did not regonize my WinXP (NTFS5) :o I had to reinstall WinXP, next time I will try LILO.
Actually, you might wanna give SuSe a try if you can, the latest version has beowulf cluster packages. Basically, Beowulf out of the box. The only problem is that you have to buy it or FTP install. THe lab I was in doesn't have a net connection, just because us CS students are naughty. ;)
Oh stupid and/or rude me.
Thanks to the two of you, Nite_Hawk and Pascal. Your words of encouragement and advice were a great help and I'm greatful. I hope to repay them someday.
nice informative thread :)
Is that supposed to be sarcastic, Panajev2001a?
I finally installed the RedHat 7.3 in an separeted SCSI drive and it is working great. I am using gcc and doing some linux system call tests with C. I will try some mutex with clone tomorrow :wink:
Please anyone know some good linux´s kernel tutorial (book or site)? thanks
I'll see what I can find you interms of material. I know where to get a lot of ebooks on it. I just don't know about legalities.
vBulletin® v3.8.6, Copyright ©2000-2013, Jelsoft Enterprises Ltd.