On Wed, 2002-09-11 at 09:47, Jim Wildman wrote:
> So you expect the latency and nondeterministic nature of IP
> communication to be solved satisfactorily enough for meaningful disk
> traffic to occur?
>
> I have not looked at in depth, but iSCSI strikes me as a round peg in a
> square hole solution.
If you've done much work with FCAL or switched Fibre Channel, you know
how much of a pain it is to troubleshoot a LIP'ing ring problem or do
just about any debugging without a Fibre Channel sniffer.
Enter gigabit ethernet: now you can use your standard IP tools (ethereal
comes to mind) to troubleshoot problems, and you avoid some of the
cross-vendor compatibility issues and HBA driver bugs by abstracting
things one step further.
An iSCSI SAN would ideally be one unified switch segment - preferrably
off of the same switch with some serious backplane speed. It's far
cheaper to work with a good gig-ether switch than the more expensive
Brocade Fibre Channel switches (which is the only real way to get Fibre
Channel working reliably anyway). You would want this segment cut off
from the outside world, and you would administer it as a SAN solution -
not as some dinky hub-and-bridge network that most IP traffic endures.
If you haven't lived through the pain of getting a Fibre Channel SAN
working in a production environment, TRUST ME, iSCSI is a Good Thing.
- Ian C. Blenke <icblenke@nks.net> <ian@blenke.com>
http://ian.blenke.com
This archive was generated by hypermail 2.1.3 : Fri Aug 01 2014 - 19:16:38 EDT