Obsoleting QAS in SPI-4

Robert Elliott relliott at hobbit.eng.hou.compaq.com
Fri Apr 28 11:35:24 PDT 2000


* From the T10 Reflector (t10 at t10.org), posted by:
* relliott at hobbit.eng.hou.compaq.com (Robert Elliott)
*
As discussed in yesterday's Parallel SCSI working group meeting,
here is the text for 00-228r0.  It should be on the web site
shortly.

I propose we make the QAS (quick arbitration and selection) protocol 
obsolete in SPI-4.

Reasons:
* Performance analysis of the fast-160 training patterns has shown 
that adding 2 microseconds is inconsequential.  QAS only promised 
to reduce arbitration time by a comparable amount.  If adding 2 
microseconds doesn't matter, subtracting 2 microseconds doesn't matter.

* It will simplify logic design and testing if QAS is not included.

* There is no fairness algorithm between QAS and non-QAS devices, 
so QAS should only be enabled when all devices on the bus support QAS.  
QAS can only be enabled when packetized protocol is enabled.  
Enclosure management devices and tape drives are unlikely to support 
QAS and packetized protocol any time soon.  Presence of these devices 
prohibits using QAS in many environments.

* QAS depends on snooping traffic from other initiator-target 
connections.  It isn't clear whether this will work through 
all expanders.  There is no domain validation technique defined to 
verify that a domain supports QAS before enabling it.

* The SCSI Trade Association (STA) has defined "Ultra 320" as 
including packetized protocol but not including QAS.


-- 
Rob Elliott    
Compaq Storage Products Division
mailto:Robert.Elliott at compaq.com

*
* 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