“It would be surprising if you haven’t read or heard about the
newest kid on the supercomputing block-Beowulf clusters. For those
still in the dark, a “cluster” is a set of computers that work in
tandem on a task. In the case of a Beowulf cluster, those computers
run Linux and can be used to attack a math problem or to even
provide high-availability file services. Beowulf clusters can be
composed of anything from Intel 486 PC’s to SPARC-based Sun
servers. The only requirement is that all of the computers be
running Linux and be located on a network.”
“For those interested in building a Beowulf cluster the only
real resources have been web-based. For example, there are several
pages hosted by NASA, the creator of Beowulf clusters, that detail
the setup and configuration of Beowulf clusters. However, these
resources rarely delve into the more mundane tasks required of a
Beowulf administrator or user. For example, have you considered
whether your computer room (or perhaps your bedroom) can actually
provide enough power for your cluster? How about whether the floor
can actually support the weight of several computers vertically
mounted? Obviously, something else was needed to address these
issues.”
“In order to address the concerns of those wishing to build
Beowulf clusters O’Reilly and Associates published Building Linux
Clusters, written by David Him Spector. The stated goal of Building
Linux Clusters is to be a “primer for the veteran Linux programmer
who wants to take the first steps into parallel computing.”
However, after reading the book you will find that the goal would
have been more appropriately stated as, “a primer for the veteran
Linux administrator who wants to take the first steps into parallel
computing” because Building Linux Clusters does not really cover
programming or algorithm design as it pertains to parallel
computing applications. Rather, the book gives a very detailed and
useful overview of what a Beowulf cluster is, how to plan one, how
to build it based on that plan, what factors to consider, and what
tools are available for use on a cluster.