Re: Command Queue Depth (was asymmetric/Symmetric)

[Stephen Bailey <steph@cs.uchicago.edu>]

> > So it sounds to me like this distribution of (command) 
> > resources across initiators is a T10 SCSI issue, and should be
> > solved there, not by each individual transport (yesterday FC -
> > which didn't solve it, today iSCSI, tomorrow IB or whatever).
> [...]
> One thing I got from Jim's note is that SCSI storage devices designs are
> carefully tuned. So, behaviors at the transport layer do make a difference
> -- especially to cost-driven products.  In my opinion, the IPS community
> would be well advised not to ignore these concerns.

It seems like poor standards practice to work-around attached
standards that are under active development.

iSCSI should engineer to the level that it does not make the problem
of handling oversubscription WORSE than on other LLPs (and can benefit
from an ultimate ULP solution), but the real fix does belong in the
SCSI layer, since it's responsible for the architecture of initiators
and targets.

It would be different if this problem were not even on the radar
screen of T10, but the annecdotes about FC RAID boxes that can't do
ANY useful work because they're too busy BUSYing incoming FC commands
are already out there.  If they aren't moving quickly enough, is the
task of interested parties to drive the T10 process along.

Render unto Caeser and all that.

Steph

[So which is Caeser?]