SBC-3 Power Conditions model

Kevin_Marks at Dell.com Kevin_Marks at Dell.com
Wed Mar 25 15:06:03 PDT 2009


* From the T10 Reflector (t10 at t10.org), posted by:
* <Kevin_Marks at Dell.com>
*
I have no issue removing the duplicated text around REQUEST SENSE
command responses, but I think there are other fundamental problems that
I as the new tar baby owner will have to fix, while enduring the Houlder
Effect. And this is mainly that most of the text talks about a command
that requires the blah blah power condition.  To my knowledge (other
than START STOP UNIT command), no other command requires a specific
power condition to complete other than ones that require the active
condition. Or put another way, can someone tell me a command (other than
SSU) that requires a standby power condition or an idle power condition?
They can be processed to completion in the standby or idle, but they do
not require theses states.
Kevin
-----Original Message-----
From: owner-t10 at t10.org [mailto:owner-t10 at t10.org] On Behalf Of
Gerry.Houlder at seagate.com
Sent: Wednesday, March 25, 2009 10:40 AM
To: t10 at t10.org
Subject: Re: SBC-3 Power Conditions model
* From the T10 Reflector (t10 at t10.org), posted by:
* Gerry.Houlder at seagate.com
*
I have added comments in-line.
	     Ralph Weber
	     <roweber at ieee.org
	     >
To 
	     Sent by:		       "'t10 at t10.org'" <t10 at t10.org>
	     owner-t10 at t10.org
cc 
	     No Phone Info
	     Available
Subject 
				       SBC-3 Power Conditions model
	     03/25/2009 08:28
	     AM
* From the T10 Reflector (t10 at t10.org), posted by:
* Ralph Weber <roweber at ieee.org>
*
I have a few questions about 4.17.1 in SBC-3.
Regarding the following entry in the first a,b,c list:
"b) the power consumed by the SCSI target device while
in the stopped power condition should be less than the
power consumed when the logical unit is in the active
power condition or any of the idle power conditions
(i.e., for devices that have a rotating medium, the
stopped power condition shall stop the rotating medium);"
Question: The absence of any mention of standby in the
text suggests that a device in one of the standby power
conditions can consume less power than a device in the
stopped power condition. Is this a desirable outcome?
>GAH: STANDBY may be equal to STOPPED, so we don't
>want to recommend STANDBY to be "greater than" STOPPED.
>I know it only says "should" but for real implementations
>on disk drives, customers will expect this to be "shall".
>Note the shall in the i.e. phrase.
>If you see the need to add a "and should be less than or
>equal to the power consumed in any of the standby conditions"
>I would not object. I can't forsee any real implementations
>that wouldn't design stopped to be the lowest possible power
>consumption of the group.
Two paragraphs farther down the subclause there is a
list that describes the responses to a REQUEST SENSE
command if a logical unit is in an idle power condition.
The list is identical to a list in SPC-4 definition of
the REQUEST SENSE command.
Questions:
Do we need the same lists in both standards?
My recollection is that past SPC revisions have omitted
discussion of REQUEST SENSE response caused by command-
induced power conditions changes, but that is not the
case in SPC-4 r18.
Do we want to go back to the former condition, or do
we want to remove the list from SBC-3?
>GAH: My recollection of the reason the list is duplicated
>in the power management clause of SBC is that SPC can make
>a specific reference to a clause in REQUEST SENSE command
>to indicate the required sense code behavior but SBC can
>only reference SPC (no specific clause) and this was deemed
>to be too obtuse a reference; thus the sense data info was
>added to the power management section of SBC. Also SBC adds
>Stopped condition options to the list; if the stopped condition
>rules were in the SPC text then perhaps the SBC text is not
>needed. I note that SPC already references SBC for START STOP
>UNIT mentions, so referencing it for Stopped condition also
>would not be much of a stretch, especially since there is now
>a glossary entry for stopped condition.
A similar problem exists with the REQUEST SENSE responses
when the logical unit is in a standby power condition,
but the plot (as they say) thickens. In this case, SPC-4
is running half-and-half. One of the command cases than
can result in a standby power condition is mentioned, but
the other is not.
>GAH: it is true that the SBC text adds cases for "entered
>by command" that aren't in the SPC text. This is also caused
>by the fact that START STOP UNIT is not part of SPC and
>non-block devices do not have those cases.
>Clearly the SBC list of cases is the complete list and SPC
>is a partial list. Is it reasonable to have only the complete
>list in one of the standards and make the other standard(s)
>reference it, even if the "other standard" has no method to
>implement some of the cases? Up until now T10 has opted to
>duplicate such rules. Note that we have 3 copies of the
>power condition model because of items only allowed within
>certain standards. This "three copies" problem is a far
>bigger problem than the "two copies" problem you cite for
>the sense data descriptions.
Surely, we will want to treat idle and standby in the
same way, and the responses to the above questions
should guide achieving that goal.
All the best,
.Ralph
*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org
*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org
*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org



More information about the T10 mailing list