After a good friend of mine from high school stop by the other day I got to thinking about a project that I had started and then put on a shelf. We were talking about computers and the systems that he had built and it got me thinking again about my Cluster Computer project so I am revisiting it.
I have been fascinated with multiprocessor systems for years not only because of the greater load of work they can handle but also because of the physical design of the entire machine. I guess I get more into the design of computers than their actual function and this is why I have some sitting around the house because they look cool not so much that they actually perform any actual function here.
The coolest looking computer in my book is the Cray T90. This thing was a gold and black computer with no wires. It weighed almost 65,000 pounds, used 400kW of electricity, and back in 1991 cost around $10 million. There was one for sale on eBay back in 2000 for $212,000 but it used proprietary software and none of the licenses were transferable so I have no idea what you would do with it if you bought it.
My aspirations have been a little more low key compared to the Cray system but still I have tried to pull off something along that concept. Initially I had looked at articles about Beowulf systems that where just clusters of any old computer you had. Then when computer hardware prices dropped enough, people were filling rooms with towers and making cluster systems that way. Nowadays people have learned that you don’t need the whole case and have just linked the motherboards in racks or racks of video cards to do data mining for virtual currencies. These are the inspiration for my creation but I have one other source that I consider my goal when designing and building my own cluster.
The Humidor Cluster by Jeffrey Stephenson, is what made me think that this was something I wanted to build. The only thing about it, and all of the other systems before it, is that they use full size motherboards which require a lot of full size components to be tucked elsewhere or crafted into the design somehow. My thought was to use something that is already built to be compact, a laptop.
So for a while I was just dreaming about this because I didn’t have the money to get four or five laptops of all the same design to start with but eventually an opportunity came along. The local schools were giving away their surplus equipment and some of those included Acer AspireOne Netbooks. They were tiny little Windows XP computers with Intel Atom CPUs. I was able to get the basic components together for a small cluster and began my venture.
The main reason for using the laptop instead of a full system was for not only compactness but also because they provided abilities much like blade servers. Each laptop comes with built-in video, USB, Flash Memory Reader, Sound, Ethernet, WiFi, cooling fan, keyboard, mouse, and a battery backup. I found out later that I had to use individual power packs for each laptop but even they are much smaller than a standard motherboard power supply. The first four machines all assembled together take up the same space as a box of paper.
First thing was to remove the screen from one of the netbooks and see if it would still boot to the OS without generating an error. The reason for this is because it allows the system to run cooler and it makes it even smaller in height thus saving more space. Removing the screen didn’t affect the system in any way and I was still able to connect a monitor to the external VGA out on the system.
After that I installed Linux Mint on all of the systems, made sure it was up to date, and then tried to figure out the clustering … and that is all the further I got with it. Linux Mint isn’t really designed to work in a cluster so I have to find another flavor that works better and some of the power packs were bad so I have to get more of them yet.
Don’t be fooled by those people that rave about how much better Linux is versus Windows. Granted, it’s free and there is a lot of settings you can use to configure it but if you have no idea on how to use it, you are just going to be just as lost as I was. With Windows or IOS, you install your application and then run it. In Linux, you have to decompress the files, build the code into an executable, and then find the file setting to change so that the application will run when and where you need it to. When it comes to Cluster Computers, this can be a little more frustrating if you are not careful in paying attention to what you are doing on each node.
So, now I’m looking at this project again and figuring out how to proceed. I think I found out how I can get the software to work by using Rocks Clusters software kit. The process is still going to be a bit of a headache because I will have to cheat on some of the installation tactics as these netbooks were not designed for this and they don’t have built-in optical drives. If I get it to work, I will have a little four-node system all to myself.
The bigger question is, what do I do with it?
I thought about making it a LiteCoin miner just for kicks. I doubt it would ever turn a profit but just to see it working would be fun in itself. Even if I knew how to code, I still really have no idea as to what good it would do me. I will probably run it for awhile and then shut it down and let it sit in the corner and if I ever get more AspireOne netbooks, I could make it bigger.
If I do sustain the system for some continued purpose, I want to see if it will run on just WiFi. That way I wouldn’t need as much cabling and I could have as many connections as the router can handle without the mess. It would be on its own channel and work group as not to interfere with the regular wireless in the house just in case the traffic were to slow down my streaming of Facebook …
I will play with it and see if I can get an update with some more pictures.