Slashdot Mirror
Xbox Linux Cluster
aeiz writes "Adam Cecchetti put together a Linux cluster using 3 modded Xboxes and was quite surprised with the results. He used the Mandrake Linux distro." illumin8 summarizes Cecchetti's conclusion after investigating "the cost-effectiveness (or lack thereof) and trouble involved in setting up a 3 node Linux cluster based on the Microsoft hardware. The end result: A cheaper Walmart PC would perform better at the same task."
Stats:
3 XBOXes
Mandrake 9
MPICH
distcc
Quick Overview:
I was walking back from class thinking to myself I haven't done much in the way of parallel computing lately. So I thought it would be a interesting idea to test the usefulness of the XBOX platform as a cluster node, as the XBOX Linux project says they are useful as. At the same time I had heard a lot on various message boards that individuals were using the XBOX as a Linux desktop to check email and browse the web, etc, etc. So, I wrote a grant proposal to Saint Vincent College asking for funds to build a Linux cluster out of XBOXes. I was issued enough funds to purchase 3 units and accompanying hardware for the project.
Parts
After ordering all the neccessary components and a stop to the local Best Buy. I had all the necessary parts to built this cluster.
3 XBOXes
3 Matrix Mod Chips
1 VGA Switch
1 USB Keyboard
1 USB Mouse
1 USB Motherboard Card
Install
My first step was to flash the mod chips, as they came without a BIOS. After some digging on the web it became clear that the only place I was going to find BIOSes were from irc channels, where they were more than happy to hand them out. The programer that came with the Matrix mod chips was easy enough to use. I simply had to choose my BIOS file, and the program wrote and verified it for me.
The alignment of the chips on the board took sometime, however compared to earlier chips that required 29 wires be soldered to the motherboard this was the only viable solution if the XBOX was to be considered as a replacement for a store PC. The next step was to get Linux onto the XBOX. I used the ISO images boot and install from the XBOX Linux projects web site. Most people on the web had luck with cdrws instead of cdrs and mine worked just fine as well.
Installing Mandrake Linux was the next step. The install disk simply wrote the image over the disk and only took about 15 mins to do so. The only problem was that all three units came with different size hard drives. The hard drive units were not clearly marked to their size, and I ended up pulling out my calculator to calculate their disk size.
After swapping disks the Mandrake partition started right up with no problems. I had no trouble using the XBOX as a standard desktop PC after the install. Nor did anything for day to day computing seem missing. Seems if you can get this up and running its fairly easy to use PC after.
Cluster
The clustering of the XBOX units would not go as easily. I wanted to be able to remote admin the XBOXes and the MPICH package needed ssh to pass commands and I was not about to use rsh. After downloading the sshd source from openssh.org, I found my next big missing piece. Gcc was not installed as well. After installing eight rpms for the gcc package I had a C compiler. The version of openssl was outdated for the current version of openssh so that had to be installed before moving on too. It would seem that the Mandrake install was minimal indeed! As even zlib, required by openssl, was not installed. Finally after a few hours of installing and searching I had a node that was ready to install my clustering software on. Distcc which would allow me to cross compile software across nodes was installed. I also installed MPICH as I had experience with it from my other projects. Finally things were ready to go.
Results
The XBOX unit functions well as a desktop computer for general usage, email, web browsing, etc. The total cost of the unit with keyboard, mouse, and parts came to 383.72 dollars. However, it is not as cost effective or as easy as using a general PC obtainable from almost anywhere for the same purpose. The technical merit of soldering the usb to XBOX controller wires as well as installing the mod chip are beyond the technical skills of most. In short if you already have an XBOX and want to tinker/need a computer and have no fear of the possibility of ruining your unit then this is a viable solution. However, as a general replacement to a desktop pc, the XBOX is not nearly as userfriendly or cost effective as a 200 dollar pc from walmart.
As for the XBOXes as a cluster I have to admit that I obtained better results than I had predicted. This was in part due to that the XBOX has a 100 Mbps Ethernet card and I was told they had only a 10 Mpbs card. I recompiled the Linux kernel 2.4.20 on a single node took 48 mins 30 sec. Using distcc the process was cut down to 20 mins. This represented a 2.4 times out of 3.0 speedup. Not to bad for the 3 nodes. My MPICH results were less telling. After nearly 100 runs of the same test programs my results varied between twice as slow across the three nodes and five times as fast as running the same program on a single node.
In conclusion, the XBOX functions well due to the work done by the XBOX Linux project. However the same or greater computing power could be obtained for an equal price, without the complications of modifying the XBOX. This makes the XBOX an unfit solution as a replacement for a personal computer or a cluster node.
Well, on the whole, fairly uninteresting.
But in case people don't know about it. distcc, the clustering solution he used rocks! It's just a wrapper for gcc, and works on most platforms. We use it here at work on Mac OSX!
http://distcc.samba.org/
super easy to set up too!
If you rtfa, like you pretended, you would have seen that he reached the same conclusions.
Quote from article:
The XBOX unit functions well as a desktop computer for general usage, email, web browsing, etc. The total cost of the unit with keyboard, mouse, and parts came to 383.72 dollars. However, it is not as cost effective or as easy as using a general PC obtainable from almost anywhere for the same purpose. The technical merit of soldering the usb to XBOX controller wires as well as installing the mod chip are beyond the technical skills of most. In short if you already have an XBOX and want to tinker/need a computer and have no fear of the possibility of ruining your unit then this is a viable solution. However, as a general replacement to a desktop pc, the XBOX is not nearly as userfriendly or cost effective as a 200 dollar pc from walmart.
And your point was, what?
---
The combined human population is enough to feed every living tiger for app. 28000 years.
Its not dead, read up on it if your going to post about it.
Release coming Feb. 15th
http://www.theneoproject.com
While he states ~$370 per Xbox, you don't need a keyboard, mouse, and mod chip for every xbox.
With the right modchip and a small bit of sodering you can flash the bios (TSOP) already on the motherboard then remove the modchip to repeat on other Xboxs. After the first xbox + supplies, setting up additional xbox is just going to cost the the xbox and a bit of time (1 hour or less). So each additional xbox should be 199.99 + tax.
Granted, it's still not very effective clustering solution.
Or perhaps they could have implemented a grammar checker for you.
My beliefs do not require that you agree with them.
There are all kinds of issues such as MPICH's underlying communication structure (Tree, Linear, Cube, whatever). The fact that he could not get his MPI test program results consistant tells me that he has no idea how to configure it.
Most likely he did the standard autoconf/make install. The code for MPICH is out there if you want to see it. Basically for TCP/IP it is all-to-all, or at least it was for the 5 years I ported it and helped maintain it. It wasn't that efficient because of other reasons.
He could have harnessed the power of the Geforce in his distcc also if he had just wired jumpers between pins 1, 4, 5, 6, 12, 14, 16, & 23 of the Geforce to pins 12, 13, 3, 16, 33, 1, 2, & 9 of the BIOS chip. This has lead to over 16% speed increases in kernel compiles on many LinuXBoxes.
details here and here.
Cheers
It's actually Saint Vincent, in Latrobe Pennsylvania (that's about 30 minutes or so from Pittsburgh); think it's a Catholic college too.
/. user something like this probably doesn't seem too complicated, or maybe even original, but for a middle-aged professor it's something new.
Anyway, you'd be surprised how easy it is to get grants like this at many universities. Sometimes it's not about how great or elaborate a proposal is, but whether or not it seems interesting to faculty. Since you're a