| As far as performance goes, the three FCAL loops are peers in the system.
Is this true in a clustered failover environment as well
Sure.
or can I tag the failover loops as having a lower priority than the "local" loops?
What exactly do you mean by "local loops"? If by "local loops" you mean FCAL loops that are private to an individual filer within the cluster, then I'm not entirely sure that we support that type of configuration anyway. In a cluster configuration, each filer requires at least two FCAL adapters, one to connect it to its primary disk array, and another to connect it to its secondary disk array which it takes over in the event of a failure (which is the primary of the other filer in the cluster) . It is "normal" for that to be the beginning and end of the story, except in the case of a large capacity cluster that requires two pairs of FCAL adapters per filer (see below). Just from the quantity perspective, the only filers that can or will be able to receive a third FCAL loop are the F630 and F760 units, so this discussion is somewhat moot in the context of an F740.
As a point of information though, the Data ONTAP 5.2 release only supports F630 filer clustering. It does include support for the "large capacity" cluster I mentioned above (4 FCAL adapters per system, but topping out at 168 drives across the 4 loops, not 224, so an "obvious" sample configuration would involve implementing 42 drives per loop, giving each *individual* system a capacity of 84 drives or ~722 GB gross, adding up to the ~144 GB gross in the whole cluster), but F700 series clustering for the F740 and F760 systems is not yet shipping. Patience... :-)
| Well yes, but distance-from-the-controller is not a significant
influencer
| of device performance on an FCAL loop. This type of fibre channel
There must be some guidelines though. I would think that you would not want to keep all the loops within some reasonable distance.
I don't think so. That's kinda like worrying about communications between the nodes at the opposite ends of an Ethernet segment being slower because the bits have further to walk! :-)
That is, I would probably want to put my filer in the middle of the rack so all loops are semi-equal as opposed to at the end of the rack (thinking that a filer that takes up 3 racks with 3 fully populated loops).
Yes, there may be some physical "can it reach" issues on a three FCAL loop system. At the moment, I know that our standard issue FCAL cable that takes an FCAL adapter output plug on a filer to the input of the first shelf is just one meter long. That's no problem on a two loop system, as you can just go taking off in opposite directions to go along the shelf chain for each loop (one chain heads off up the rack, the other down, that kind of thing...) but I guess it does get kinda interesting when you've got three loops to worry about. I'll have to leave this one to someone at NetApp who knows exactly how we recommend this work physically.
Indeed. fcal is quite spiffy.
It is a thing of beauty. :-)
Actually, why the decision to use copper instead of fiber?
I think the reasons were plentyfold. Affordabality was certainly in the mix.
Keith