[Gllug] RAID Partitions

John Hearns john.hearns at clustervision.com
Mon May 10 12:09:00 UTC 2004


On Mon, 2004-05-10 at 12:38, Richard Jones wrote:

> 
> Sounds slow!  Do you have a specially reserved network to take this
> traffic?  How does it handle network failures?
We're just using the standard gigabit interface.

In this application, we are using it for failover.
So data is only written when changes are being made to the disk.
I did a lot of work in this case in moving the Sun Gridengine spooling
directories to local directories on the nodes, not NFS-mounted.

Can't give you a benchmark I'm afraid - its a production system at the
moment.

After a network failure the secondary would re-sync.

> At one job I had, I had to write a network driver which worked over
> SCSI - kind of the same thing but in reverse :-) I forget exactly why
> I did this (it was over 10 years ago), but I suspect it was because
> SCSI was a lot faster than AUI ethernet in those days, and they needed
> the speed.
And you can run networking over fibrechannel too.
(Actually, if I'm not wrong, thats the physical layer for gigabit
ethernet).



-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug




More information about the GLLUG mailing list