From: Elliott, Robert (Server Storage) [mailto:[EMAIL PROTECTED]]
Sent: Monday, October 14, 2002 10:20 AM
To: Mclean, Pete
Cc: [EMAIL PROTECTED]
Subject: RE: ATA terminology in ATA/ATAPI-7
--
Rob Elliott, [EMAIL PROTECTED]
Industry Standard Server Storage
Advanced Technology
Hewlett-Packard
-----Original Message-----
From: Mclean, Pete [mailto:[EMAIL PROTECTED]]
Sent: Monday, October 14, 2002 10:56 AM
To: Elliott, Robert (Server Storage); Mclean, Pete
Cc: Evans, Mark
Subject: RE: ATA terminology in ATA/ATAPI-7Rob,Sorry I'm so long responding, I was on vacation last week.T13 meets next week and I will bring up your suggestion.Regards, Pete----Original Message-----
From: Elliott, Robert (Server Storage) [mailto:[EMAIL PROTECTED]]
Sent: Tuesday, October 08, 2002 1:00 PM
To: [EMAIL PROTECTED]
Cc: [EMAIL PROTECTED]
Subject: ATA terminology in ATA/ATAPI-7I notice in ATA/ATAPI-7 that T13 is splitting ATA into upper level and lower level documents. This seems to fit well with Serial ATA and Serial Attached SCSI (SAS), which are both able to serve as a lower levels in that division:
Upper level:
a) ATA/ATAPI-7 V1Lower levels:
a) Parallel ATA (ATA/ATAPI-7 V2)
b) Serial ATA (SATA)
c) Serial Attached SCSI (SAS) Serial ATA Tunneling Protocol (STP)However, we run into a number of terminology problems in SAS, since SAS needs to use both ATA and SCSI terminology.
a) ATA uses "device" for a disk drive, while SCSI uses "device" for an initiator and/or target and "target" for a disk drive.b) ATA and SATA tend to use "host system" or "host adapter" where SCSI uses "initiator"
c) SCSI devices have one or more SCSI ports; ATA implies single-portedness everywhere.
d) SCSI domains can have more than one initiator and one target; ATA implies one host to one or two disk drives, SATA is one-to-one, and I understand SATA II is looking into one-to-N kind of support. ATA and SATA don't seem to have a name for "domain".
SAS (sas-r02a - http://www.t10.org/new.htm) creates some ATA terms modeled after SCSI terms to live with these conflicts:
ATA domain (initiator(s) and target(s) and expanders)
ATA device (initiator and/or target to us)
ATA initiator device (classically called a host)
ATA target device (classically called a "device")
ATA initiator port (classically called a host)
ATA target port (classically called a "device")
For ATA/ATAPI-7, would you be interested in adopting this kind of terminology? It would help our SAS mappings and would help ATA fit better into additional transports. It would also help the SCSI MMC command set which is mostly used with ATA/ATAPI but must use SCSI rather than ATA terminology. The biggest change would be changing "device" to "target device". The benefit is you can then use "ATA device" to represent any kind of ATA entity, be it HBA or disk drive. Then, "ATA" can be used regardless of whether the underlying transport is Serial ATA or SAS STP, too.
In SCSI we let the Parallel SCSI lower level continue use of "SCSI" in its own self-references, and I'd recommend you let the Parallel ATA lower level continue assuming "ATA" is its own too (no need to change ATA to PATA in the low level; just make sure SATA, SAS STP, and other new standards use new terms).
We have a UML (unified modeling language) object diagram (attached) defining the relationships of our SAS objects and showing how it maps into SCSI and ATA object models (SCSI Architecture Model - 3 will include UML object diagrams defining SCSI domain, device, port, initiator device, target device, initiator port, target port, etc.). For ATA/ATAPI-7 would you be interested in creating a UML object model?
<<Visio-ch4 figure 4 SAS object model.pdf>>
--
Rob Elliott, [EMAIL PROTECTED]
Industry Standard Server Storage Advanced Technology
Hewlett-Packard
