|
[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]
|
Re: [PLUG] distributing work
|
On Mon, 5 Jan 2004, Jeff Abrahamson wrote:
> I can think of some ad hoc ways to manage distributing the work, but
> I'm wondering if there's a simple solution that others use.
One machine running the distribution of your choice acts as the master
node. This is the one you log into. Install OpenMOSIX on this node.
On the other nodes, boot from a ClusterKnoppix live CD. This already has
OpenMOSIX installed. It's CRITICAL that you match up your kernel & OM
versions on all nodes. Aside from that distros don't have to match.
If all nodes are on the same subnet they will discover each other
automagically. Otherwise there is a simple config file you'll need to
edit to define the cluster on each node.
All of the nodes have direct access to each others filesystems already,
without the need for NFS. You start a process on one node, and it
automatically migrates to another. There are commands you can use to
immediately migrate a job on startup, with different commands set to
optimize the migration for different key parameters (i.e. I need a node
with fast I/O, or I need a node with fast CPU, etc.)
One process still runs on one CPU. But if your job is split up into
multiple processes, it should distribute across the OM cluster. Threads
won't migrate away from their parent.
There are some userland tools available to manage your cluster which you
can get from the OpenMOSIX site.
NOTE: Qt apps break horribly on OM. Forget running KDE on an OM cluster.
___________________________________________________________________________
Philadelphia Linux Users Group -- http://www.phillylinux.org
Announcements - http://lists.phillylinux.org/mailman/listinfo/plug-announce
General Discussion -- http://lists.phillylinux.org/mailman/listinfo/plug
|
|