RE: Towards Consensus on TCP Connections

[Robert Snively <rsnively@Brocade.COM>]

David,

Thank you for your calming words.

At present, all the disk arrays that attach
using Fibre Channel use exactly the same protocol that is used
by the individual disk drives, i.e. FCP (or FCP-2).  Some optional
choices may be different (e.g., if more than one disk drive is
attached, it is likely to use FC-AL-2 instead of the point-to-point
FC transfer rules), but the protocol is identical.  Disk arrays may have
a slightly larger command set, including portions of SSC-2 and
SES.  Both devices may have out-of-band command capability as well,
including SFF-8067 for disk drives and various management interfaces
for RAID boxes.  But the protocol is the same (FCP).  Disk drives
tend to have symmetrical behaviors across their multiple ports, while
RAID boxes may have either symmetrical or asymmetrical behaviors.  The
protocol is still the same.  Tape drives and storage subsystems may
have ordering requirements.  The protocol is still the same.  The splendor
of this is that it does not matter whether you are controller attached, bridge attached, or directly attached, no changes need to be made
in the protocol.

Similar behavior is true of parallel SCSI RAID boxes and disk drives.

For this and a bunch of other reasons, we have been encouraging the
use of the common packet structure and architecture implemented 
by packetized parallel SCSI, SBP-2, SSA, and FCP in the iSCSI.

Bob

Bob Snively
Brocade Communications           Phone  408 487 8135
1901 Guadalupe Parkway
San Jose, CA 95131               Email   rsnively@brocade.com


>  -----Original Message-----
>  From: Black_David@emc.com [mailto:Black_David@emc.com]
>  Sent: Friday, August 11, 2000 6:08 PM
>  To: ips@ece.cmu.edu
>  Subject: RE: Towards Consensus on TCP Connections
>  
>  
>  > I do not think that such contentious statements are useful.
>  
>  Can we calm down a bit -- or at least leave refereeing attempts
>  to folks like the co-chairs and ADs?  Thanks in advance.
>  
>  It is the case that Douglas Otis's statement to which John Hufferd
>  objected is seriously at odds with industry practice --
>  
>  > If you wish to use this interface at a controller, it should
>  > assume an identical role to a device.
>  
>  Even disk arrays that are all-Fibre Channel are often using FC-SW
>  (switched) to talk to servers and FC-AL (arbitrated loop) to 
>  talk to drives.
>  There are disk array products with Fibre Channel interfaces 
>  to servers
>  that use SCSI and even SSA drives.  Most mainframes still use ESCON
>  as their storage interface, even though there hasn't been an ESCON
>  disk drive produced for many years.
>  
>  So, I think the onus is on Douglas to explain why uniformity 
>  of controller
>  and drive interfaces should be required, given the large 
>  number of product
>  engineering decisions that have gone in the other direction. 
>   After all,
>  this
>  is the Internet *Engineering* Task Force :-).
>  
>  The overall discussion of whether the iSCSI protocol 
>  specification should
>  include both arrays and individual drives is definitely 
>  germane and in
>  scope for this list - there are words in the draft charter 
>  to that effect.
>  
>  --David  
>  
>  ---------------------------------------------------
>  David L. Black, Senior Technologist
>  EMC Corporation, 42 South St., Hopkinton, MA  01748
>  +1 (508) 435-1000 x75140, FAX: +1 (508) 497-6909
>  black_david@emc.com  Cellular: +1 (978) 394-7754
>  ---------------------------------------------------
>  
>