Beowulf vs OpenMosix or anything else.

[chatgris]

Well, right now I have a data engine that handles large email lists.. This app currently stores all the email addresses permanently in an AVL tree in RAM mirrored on disk, with extra data on disk.

There's more to the app (I don't actually keep the emails in ram all the time, I cache etc but for pusposes of this question, most of the work I do I use data in RAM).

Anyways, I need to port this application to work on mutliple servers... Of course the simplest way to do this is just to split the data up with roughly x emails per server etc... and distribute processing that way.

However, I've been trying to find out information on clusters... and so far it seems that the only thing that clusters do is share CPU power (fair enough).

So basically, I will have data across multiple servers (each "list of emails" stores itself on one computer) and I would like to share the server combined CPU power to crunch through the jobs.

My questions are as follows (if you know even one it'll help me =).

1) Do beowulf clusters outperform OPenMosix clusters? (I've been tyring to find information on beowulf clusters and their special required libraries but to no avail)

2) Where can I find information on beowulf clusters?

3) If beowulf clusters have better performance (which I am guessing that they do) how difficult is it to port a standard data crunching daemon to a beowulf cluster? I can perform my task using threads or processes, no big deal, but what kind of gotchas await me?

I know I've asked a lot here, but any information, even just some links that could give me more insight into these technologies (especially beowulf, I haven't been able to find out anything about that) Would be greatly appreciated.
_________________
Open your mind. Open your source.

Due credit for avatar from http://www.aikida.net

[AlterEgo]

[offtopic]
Regarding your signature: http://koti.mbnet.fi/~hoppq/sc-howto.html
[/offtopic]