[Probably OT]: Data replication

[Greg Lindahl]

> I've read this mailing list for some weeks and I recognized I probably
misunderstood
> the meaning of cluster. This mail is to better understand what this word
means and
> to ask something I really and absolutely need.

That's because there are several definitions.

One kind is "high availability", where you have a cluster in order to
increase availability of a resource. Your example of 2 NT servers falls into
that category.

One kind is "high performance", where you use many machines on a single
program, usually parallelized using MPI. That's what most beowulf people do.

A third kind is "high throughput", where you use many machines on many
programs, which usually don't have to be parallelized. That's what sites
like Incyte and Google do (thousands of nodes each.) Some beowulfs are used
this way.

There are some good webpages for Linux high availability, and TurboLinux and
RedHat sell special software for this.

-- g