Please forgive me if this is totally newbish, but I haven't a clue about linux clusters. I've looked around (briefly) on the web and could not find something as simple as an explaination of what a linux cluster even is. There are plenty of tutorials for setting up linux clusters, but I felt those would be more info than I needed at this point. I'm really new to the concept of clustering and would like answer to simple questions like: What is a linux cluster? Do all the machines function as one? Is a cluster like a multi-processor machine but split up over a network? Can I just take a bunch of linux machines, network them, and "have" a cluster? Do the machines in a cluster share a single HD (would make booting an issue, perhaps, but what do I know!)? I'm totally in the dark about distributed computing, any light is helpful on this as well!

As you can tell, I'm completely clueless. Thanks for any help.

A cluster is basic'ly a bunch of computers networked together, and seemingly working as a single unit. There is one master computer, that the user interacts with, and a number of nodes. The master computer delegates work to the nodes. The nodes aren't typically seen by the user, because they are usually connected to the master on a separate, sometimes really-high speed, network.

The master computer usually has a full install of the OS. The nodes might have a small hard-drive and minimal install. They might also network-boot and get the OS image from the master, and run everything in memory.

How well a cluster works depends a lot on how well a program is written for it. The better a program is written, the more it is able to be split up to work on the nodes. If it's written poorly, it won't be run on as many nodes as it could, or the processes might collide and waste time, or there might be problems gathering the results.

There are clustering systems that sound like they could work pretty good for the hobbyist. You network your computers, install some software, and they can compare work-loads and abilities, and pass around processes to the computer better able to handle it. Say you are working on configuring a kernel on your P2 notebook. You start it compiling, and the system looks around your network and finds your Athlon XP desktop sitting idle. It passes the compile to the Athlon, and you can keep working on the notebook.

As for just cobbling a handful of old 486s together, it won't run like a super-computer with just day-to-day software. Few regular programs take advantage of dual-processors, let alone clusters.