I agree with you and Todd but I've another point of view. I know very well other competitor (really ones) in the field of 'cluster' such as Isilon (and this one considers a real competitor only NetApp). For Isilon and such companies cluster mean over all HPC, parallel f/s, single name spaces, grid computing, they mean also to share network and other resources in a scale-out point of view.
I think for that reason that to propose CM to justify a 'never poweroff', 'move a single vol without interruption' and so on is a very limited vision maybe because in the scale-out sharing of resource the NetApp disk ownership is a constraint? Or more realistically because to speak of CM thinking to 2200 systems could be a nonsense (customer with limited budget and needs..
regards
-----Messaggio originale----- Da: Bennett Todd [mailto:bent@latency.net] Per conto di Bennett Todd Inviato: domenica 17 giugno 2012 18:27 A: Blackmor, Chris; Milazzo Giacomo Cc: toasters@teaparty.net Oggetto: Re: Cluster mode - Market vs field
On Jun 17, 2012 12:08 PM, "Blackmor, Chris" Chris.Blackmor@amd.com wrote:
With all the missteps NA has had getting C-mode out the door I don't think they can really push it on anyone. Remember Auspex 1.8. Where's Auspex now?
A very fond memory, if triste; the months that felt like years that I spent trying to master the art of handholding the recovery of lock state after failover, and appreciating the cleverness of the multicast tricks and nfs protocol.shortcuts, taught me lessons about how NFS really works (when it does) that will last me.
H-A is great when it works, but I've rarely seen implementations that aren't seriously lower availability than the single-point-of-failure system they try to replace.
Though I think what did for Auspex wasn't their ambitious H-A, it was NetApp's ability to cling tenatiously to the Juggernaut as Moore's Law drove costs of processing, memory, and bandwidth down, and down, and ever down.