2008-08-29 07:43 PM
I'm going to put my money on NFS or other similarly network driven protocols.
I'm not knocking the other protocols - however when I think about something which might need to explode and scale at a moments notice - provisioning fixed sizes with fixed paths with (inherited from the scsi protocol) fixed boundaries in the form of Queue depths and a number of other boundaries.
Even with iSCSI running over 10G, we're still bound by SCSI and the limitations it brings to the table when you start to tax it and try to cripple and destroy it.
But then looking at costs associated with Cloud computing... The internet did not grow overnight to create some of the clouds we work with, and the point at which it exploded it was because of open systems, open protocols and the ability for "anyone" to do it on a shoe string budget.
So, while the cost efficiency and simplicty of iSCSI is a formidable player, and FCP can bring its weight in gold and performance - I don't see either of those two protocol sets being able to scale at the needs, capacity, growth and expectations at a pace with which we're accustomed to.
Much of it will depend upon who win's the cloud - Blocks or Files, what type of workload we can expect (or not expect) and where it runs with.
... But I'm just tired after driving for 10 hours today, so my mind could completely change in light of new evidence!
I'll have to think about this some more Vijay! Thanks for the inquiry!