Question Regarding 09-268 SPL - STP Extended Buffer Support
Tim_Symons at pmc-sierra.com
Mon Nov 2 11:48:12 PST 2009
* From the T10 Reflector (t10 at t10.org), posted by:
* "Tim Symons" <Tim_Symons at pmc-sierra.com>
Outline for why we need the propagation delay figure:
SATA uses HOLD and HOLDA to throttle data transfers that works fine for
direct connected short cables, but requires additional buffering when
expanders are used in the connection path.
SAS uses XRDY to issue credits for flow control and does not have this
To allow support for long cables in SAS topologies there needs to be a
mechanism to ensure correct operation of SATA.
09-268 is a proposal that determines the amount of buffering required to
support SATA connectivity over long cables and a method for disabling
SATA protocol support if insufficient buffering is available in the
A value for the Total Cable assembly propagation delay time in nS is
required to determine the number of data words "in-flight" that needs to
be buffered by the expander.
Proposal 09-268 section 18.104.22.168.3 (STP flow control buffer size) should
be reviewed for this discussion.
Using absolute values the following equation is used to determine the
buffer size (in dwords)
Minimum buffer size = [20 + ((2 x Pd x R) / 40)],
Pd = Propagation delay of cable assembly (ns) (see SAS-2.1)
R = Logical link rate (Gb/s)
Minimum buffer size value shall be rounded to the next highest integer
The propagation delay of the cable assembly shall be measured at the
connector i.e. cable delay + active component delay
If any rounding is to be done for Pd (nS) then this shall be rounded-up
to the next integer value.
This calculation determines the MINIMUM buffer size.
Principal Engineer, PMC-Sierra Ltd.
Cell: 778 998 5025
E-mail Tim_Symons at pmc-sierra.com
From: Olawsky, Barry [mailto:Barry.Olawsky at hp.com]
Sent: Monday, November 02, 2009 9:30 AM
To: Tim Symons
Cc: t10 at t10.org
Subject: Question Regarding 09-268 SPL - STP Extended Buffer Support
During the SFF-8636/8449 conference call on October 22nd a question was
posed for which we need your assistance to answer. Tom Palkert of
Luxtera proposed using cable length to estimate the end-device to
end-device propagation delay. To determine if this is a feasible
technique, we need to know the maximum acceptable total propagation
delay error including resolution and accuracy errors. Since you propose
limits on this parameter in 09-268 we thought you might have valuable
input on this topic.
* 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