Clustering with FreeBSD

Andy Sporner sporner at
Tue Feb 10 04:20:09 PST 2004


It is a different subject and I sort of hate threads that are misleading.

So far it hasn't happened yet, but before it can.  :-)  The clustering that
I am offering is *NOT* Beowulf-like.  It is more geared towards
Internet Application HA.    In other words--server X dies, what should
server Y and Z do to make sure that the stuff on server X does not have
to wait for server X to recover.

Somewhere else on my site I have a utility called FREP.  In my test
area in my lab I have the two things integrated.  

There is in Linux-Land a thing called sometime like "Remote raw
disk" (can't remember specifically what it is called) but it gives a
local device node for a remote device on another machine.

What FREP does (at the moment only in the lab) is to syncronize
access to directories and replicate the changes done by the nodes.
The idea is to be able to have a 2-3 nodes running mail servers with
the spool directories replicated (locking is on the file basis).  A
load balancer goes on the front and with this you have a scalabale
Mail server that is fault resiliant.

A lot of people in the academic community are worried about
Beowulf and for correct reason, but there is an often neglected area
which is where Micro$oft is winning in the moment and that is
in the business end of the house.


More information about the freebsd-cluster mailing list