Behavior of TEST UNIT READY command in idle and standby power conditions

Mark_Hollinger at Mark_Hollinger at
Wed Nov 28 07:27:48 PST 2012

* From the T10 Reflector (t10 at, posted by:
* <Mark_Hollinger at>
> I mean, just suppose the Target transitions to Low Power in the narrow
> time window before the last Test Unit Ready or Request Sense Poll and
> the arrival of a Read Cdb. Well then the Read must handle that condition.
Ah, but if the transition into low-power state is based on idle time, then
the TUR will reset the timer -- or at least extend the time slightly.  A
sophisticated target might go to medium-power state after 5 minutes with no
actual I/O activity, and then low-power state after 30 minutes with no
activity at all.
> The simplest kind of Scsi, the most power efficient, would forbid the
> host poll on these inescapably logical grounds.
Actually, there is a logical escape if you consider the power efficiency of
the system as a whole.	There are many occasions when the host may want to
have a look around -- survey the landscape of attached devices without
necessarily doing any I/O.  For example, at OS boot time, or upon opening a
dialog box to select an input file, well-designed software will build a list
of drives using whatever information is immediately available.	You don't
want to spin up all the drives (introducing a 5 second delay and consuming
power), but if a LU happens to be ready, you might check the disk label or
read the top-level directory.
Power-conscious background tasks such as an indexing process or a virus scan
might also want to limit their I/O activity to SCSI devices that are already
spun up.  Polling improves efficiency if used to determine whether or not to
initiate I/O to a particular device, and it's important for a target to
return descriptive information when polled.
      - Mark Hollinger
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at

More information about the T10 mailing list