Several ZBC proposals

Joe Breher Joe.Breher at hgst.com
Mon Jul 7 10:03:27 PDT 2014


Formatted message: <a href="http://www.t10.org/cgi-bin/ac.pl?t=r&f=r1407077_f.htm">HTML-formatted message</a>

Thanks for your comments Gerry. Please see below.
Joe Breher
(478) 2-Breher
(478) 227-3437
This e-mail may contain confidential or legally privileged information of
HGST. If you are not the intended recipient, please notify us immediately by
responding to this e-mail and then deleting it from your system.
On Jul 7, 2014, at 10:03 AM, Gerry Houlder wrote:
Your proposal 14-184 adds a "disallow unaligned writes" capability but
restricts it to one type of device and one type of zone.
If a device desires to enforce this behavior, it would want to enforce it
over all zone types within that device, not just conventional zones. it
should be written in a way that the alignment is enforced on all zone types
within the device.
It is restricted to host managed devices, as this is a new device type. To
make this available to host aware devices would lead to such host aware
devices being incompatible with legacy systems. SBC-4 already contains a
'should not' restriction on unaligned accesses, but it is not a 'shall not'.
Host managed devices can have either of two zone types: sequential write
required, and conventional. Sequential write required zones already prohibit
unaligned accesses. This proposal adds the option of disallowing unaligned
accesses for the other (conventional) zone type.
Also, this behavior should be permitted for all device types so either host
managed zoned block devices or host aware zoned block devices are allowed to
implement this feature and and bit in the VPD page that indicates the
behavior.
I would not be opposed to that. I would even heartily endorse that. But what
about the Prime Directive of ensuring host aware devices be compatible with
legacy systems?
This proposal may also open the door for folks that would like to reject
unaligned writes on existing direct access devices. Those folks would like to
have the bit located in VPD page B1h (block device characteristics) page so
that devices that are not zoned block devices may use this capability. In
this eventuality, the entire proposal would move to SBC-4 and not be
restricted to zoned block devices.
This is not my proposal, but the discussion may have some merit.
On Mon, Jul 7, 2014 at 10:36 AM, Joe Breher
<Joe.Breher at hgst.com> wrote:
One of my collaborators has caught that I have mischaracterized one of the
below-listed proposals - in it's very title. I will include a change to this
in r1. Please see below for correction.
Joe Breher
(478) 2-Breher
(478) 227-3437
This e-mail may contain confidential or legally privileged information of
HGST. If you are not the intended recipient, please notify us immediately by
responding to this e-mail and then deleting it from your system.
On Jul 7, 2014, at 2:37 AM, Joe Breher wrote:
* From the T10 Reflector (t10 at t10.org<mailto:t10 at t10.org>), posted by:
* Joe Breher <Joe.Breher at hgst.com>
*
I have tendered several proposals for consideration under ZBC. I'd love to be
able to present these in today's call.
14-181R0 2014/07/07 ZBC Discussion: Active zones and host managed pinning
- a view of the transitions between zone conditions, and a means for hosts to
control where latencies due to resource management are incurred
14-182R0 2014/07/07 ZBC: Read across sequential write required zone boundary
14-183R0 2014/07/07 ZBC: Read unwritten sequential write required zone
- two related optional behavioral changes with indications thereof
14-184R0 2014/07/07 ZBC: Disallow unaligned writes for host aware devices
Should be: "ZBC: Disallow unaligned writes for host aware managed devices"
- why must we support unaligned accesses on conventional zones on host
managed zoned block devices? Just say no.
Thanks -
Joe Breher
(478) 2-Breher
(478) 227-3437
This e-mail may contain confidential or legally privileged information of
HGST. If you are not the intended recipient, please notify us immediately by
responding to this e-mail and then deleting it from your system.
*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to
majordomo at t10.org<mailto:majordomo at t10.org>



More information about the T10 mailing list