SCSI Power Management
James Smart USG
smart at zk3.dec.com
Thu Jun 6 05:52:16 PDT 1996
* From the SCSI Reflector, posted by:
* James Smart USG <smart at zk3.dec.com>
>From an OS perspective, I would prefer that the drive did go back to the
ACTIVE state and complete the READ. However - only if it could do so in
a quick time period. If it takes 10-30 seconds to complete this transition
(I've seen drives take this long), then it will probably exceed the
host's timer for the READ request. The error recovery paths for the
READ timeout have to potential to be much more harsh than those that
could be kicked off due to the Check Condition response.
>* From the SCSI Reflector, posted by:
>* "Jim McGrath" <jmcgrath at qntm.com>
> Subject: Time: 7:32 PM
> OFFICE MEMO SCSI Power Management Date: 6/5/96
>The current draft SCSI specification calls for devices to return a CHECK
>CONDITION to the host if a command is issued to the device when the
>device is unable to execute the command due to being in a low power
>mode. Thus if the host issues a READ that requires disk access, and the
>is in STANDBY mode (motor turned off), it is suppose to fail the command.
>This seems to be unnecessaryily harsh and not workable. If the device
>was put into STANDBY due the the STANDBY counter expiring, then the
>host did not necessary know that the device was in a low power mode
>to being with. This will probably confuse the host and cause all sorts of
>problems. In ATA, the device will under these conditions resume the
>ACTIVE state (i.e. spin up) and execute the command. That appears to be
>much better. Moreover, desktop systems will want AT and SCSI to behave
>the same no matter what.
>Does this maek sense? ( have not worked on the power management stuff
>for some time). Comments?
James Smart email: smart at zk3.dec.com
Alpha UNIX Systems - Digital Equipment Corp. Phone: 603-881-2472
110 Spit Brook Road, Nashua, NH 03060 Fax : 603-881-2257
Disclaimer: All statements and opinions are my own.
More information about the T10