[STA Members] 6 G discussion

Alice Tate alice at scsita.org
Mon Aug 7 09:10:46 PDT 2006


STA Members & marcom team,

Following are the notes from last week's discussion on 6 G. Marty 
Czekalski (Seagate) will be revising the presentation discussed on the 
call and circulating it before the next call. Details on the next call 
are:

Next call is scheduled for:
Date: Tuesday, August 15
Time: 10 AM pacific
Tel.: 1-877-213-4847
Passcode: 781 407 0288


STA 6 G considerations discussion via telecon
August 1, 2006
10 AM pacific
 
Linus Wong, Adaptec
Paul Mutschler, Catalyst
Tom Hammond-Doel, Emulex
Yamini Shastry, Finisar
Steve Wong, Finisar
Mike Fitzpatrick, Fujtisu
Don Jeanette, Fujitsu CPA
Jim Wong, Hitachi GST
Tonya Comer, HP
Richard Tomaszewski, HP
Schelto Van Doorn, Intel
Ahmad Zamer, Intel
Schlomi Krapner, LeCroy
Harry Mason, LSI Logic
David So, LSI Logic
Paul Wassenberg, Marvell
Rachelle Trent, PMC-Sierra
Marty Czekalski, Seagate
Martin Or, Sierra Logic
Alice Tate, STA
Ashlie Fan, Tyco Electonrics
Jim Scott, Vitesse

 
Presentation was circulated to the members & marketing reflector with 
the call reminder.
 
Comments had been sent from a few member reps after reviewing the 
slides prior to the call.
 
Harry noted the goals of the meeting and what we are trying to 
accomplish. Members have been invited to give input.
 
HP reviewed their extensive comments to the slides which they sent out 
to the reflectors before the call. Comments were:
 
Slide 3: HP also wants SAS-2 HBA and expander phys to support SATA 3 
Gbps "Gen2i" drives over SAS backplanes with no port selectors.
 
Slide 4: Looking forward, SAS 12 Gbps doesn't have to be interoperable 
with SAS 1.5 Gbps if needed.  It's desired but could be dropped if it 
creates problems.
 
Slide 4: "0.5 meter backplane" is not long enough.  The OIF-CEI-6G-LR 
standard on which we are basing SAS 6 Gbps supports 1 m with up to two 
connectors.
 
Slide 4: 10 m cable is our goal, not just 6 m.  No additional cost is 
acceptable (and doesn't seem necessary).
 
Slide 5: HP has settled on multiplexing as the method for bandwidth 
aggregation.  6 Gbps to two 3 Gbps and 3 to two 1.5 Gbps are logically 
the same - the "Desired" row is not an issue.  We're willing to drop 
discussion of 4:1 multiplexing ("supporting more than two phy 
connections...".  It is only required on some links (particularly 
external links from HBA to expander, or expander to expander).
 
Slide 6: 10^-15 is our goal.  It is a statistical number, not a 
measured number.
 
Slide 6: SSC is a requirement.  All new devices must support receiving 
SSC.  It can be left up to the system vendor to decide if devices are 
configured to transmit SSC or not.  IBM seems to be the main one 
complaining about SSC in T10; they're not currently a STA member. 
 
Slide 6: SSC is also needed in 3 Gbps SAS drives - it's not tied to 6 
Gbps.  This is one of the relatively minor non-6 Gbps features that 
could cause a drive to be considered a SAS-2 drive:
* Spread Spectrum Clocking
* BREAK_REPLY primitive
* I_T NEXUS RESET task management function
* NOTIFY (POWER FAILURE EXPECTED) primitive
* BROADCAST (ASYNCHRONOUS EVENT) primitive
* RESPONSE frame RETRY DELAY field
* IDENTIFY address frame DEVICE NAME field
* Redundant primitive decode changes
* Phy test patterns
* Phy event information counters
* full active/active (data/data) ability for the ports on dual-ported 
disk drives
* full use of full-duplex in a port
* transport layer retries (maybe)
 
Overall: STA should decide on some finalized marketing names for the 
new SAS features.   Is "multiplexing" the right term, or is "time 
division multiplexing (TDM)" better?  How about a name for the package 
of non-6 Gbps features?
 
Richard reviewed some of the issues they saw but there was other 
feedback around the issues.
 
Specifically, there were SSC and PHY concerns with HP’s proposal.

Harry had to drop off the call but noted that we need to have a list of 
questions and revise the slides according to the feedback, and then 
continue on with this discussion at the September meeting in Nashua.
 
Intel noted that they will have a proposal later and that they do not 
see multiplexing as necessary.
 
Bill Ham had also circulated comments prior to the call as he could not 
make it. Those comments were: "I have serious objection to the same 
interface requirements for both disks and external interconnect.  This 
provides little benefit and seriously impedes optimization for the two 
very different environments.  Other than that this seems reasonable."
  
We have another call in two weeks (August 15) at the same time. The 
participants on the call resolved to further this discussion in Nashua.
 
Mike Fitz will discuss requirements with Alvin Cox and see if he has 
input or can participate in the next call. Alvin runs the weekly PHY 
meeting.
 
Marty is the “keeper of the slides” and will circulate an updated set 
of slides before the Aug. 15 call, using feedback from this call.
 
It was suggested that we have a evening meeting at the Nashua T11 
meeting week (separate from the normal STA meetings) to go over this, 
in order to allow participation from the T11 folks. We will discuss 
this on the next call.
 
Next call is scheduled for:
Date: Tuesday, August 15
Time: 10 AM pacific
Tel.: 1-877-213-4847
Passcode: 781 407 0288





-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: text/enriched
Size: 5593 bytes
Desc: not available
Url : http://lists.scsita.org/pipermail/scsita-members/attachments/20060807/1ae4ec7c/attachment.bin 


More information about the SCSITA-members mailing list