RE: Initiator port groups????

["Ostrowski, Stephen" <Stephen.Ostrowski@emulex.com>]

John,

	I missed port groups are dynamic across boots.  I think that there
MIGHT be a need for static mapping (or preference) for configuration.  
	Example: I have an initiator with two HBAs.  If the connection is
made from HBA 1 to a certain target, then the connection would be secure,
but if the connection if mode from HBA 2 to that target, then it would be
insecure.  So I would want to be able to configure that, and then to verify
that the correct connection was being made during login.
	Also, I could also see the need to try to balance the traffic load
across the HBAs in a non random way. 

- Stephen
 

-----Original Message-----
From: John Hufferd [mailto:hufferd@us.ibm.com] 
Sent: Thursday, December 20, 2001 4:19 PM
To: Ostrowski, Stephen
Cc: Ips (ips@ece.cmu.edu)
Subject: Re: Initiator port groups????


The direction you are going, is not the direction we intended to take
iSCSI.  The Authentication and Masking should be done on an Initiator Node
Name value, NOT based on an HBA.  There may be multiple HBAs involved in a
single, iSCSI Session.  Also there might be more then one initiator in a
Initiator Node (perhaps each with more then one HBA), I do not see why
there would need to be management beyond the Node Name.  The Authentication
is suppose to be done on a Node Name Basis.  In any event the Initiator
does not have Portal Group Tags, and Target Portal Group tags are not
required to be Static from Target Startup to Target Startup.  This
Discovery process is suppose to Discover the current values of these items.
They are NOT part of the NodeName.  Also, the Node Names are suppose to be
Long lasting Names, portal groups have no such need.  So it is a twist to
bring both together and call that an iSCSI Name.

Now, if you just intend to do mapping for the purpose of rendering a
picture of the network and connections, I suggest you look at the iSCSI MIB
Drafts and see how that might be useful or the iSCSI Technical Work Group
in SNIA, which is defining APIs for the HBAs.

In other words you need to find another way to do what you want.

.
.
.
John L. Hufferd
Senior Technical Staff Member (STSM)
IBM/SSG San Jose Ca
Main Office (408) 256-0403, Tie: 276-0403,  eFax: (408) 904-4688
Home Office (408) 997-6136, Cell: (408) 499-9702
Internet address: hufferd@us.ibm.com


"Ostrowski, Stephen" <Stephen.Ostrowski@emulex.com>@ece.cmu.edu on
12/20/2001 07:31:30 AM

Sent by:    owner-ips@ece.cmu.edu


To:    "Ips (ips@ece.cmu.edu)" <ips@ece.cmu.edu>
cc:
Subject:    Initiator port groups????





I would like to suggest a change in the initiator name. I would like to be
able to add an optional port group tag number at the end.

Example:

OLD: iqn.2001-03.com.service-provider.users.customer235.host90

NEW: iqn.2001-03.com.service-provider.users.customer235.host90,10

OLD: eui.02004567A425678D

NEW: eui.02004567A425678D,11





The problem that I am trying to solve is how to build a San Management
Platform to do the initiator to target mapping and once a mapping has
occurred, how to make sure the right connection is being made.



For example: I have an initiator with 3 HBAs, and a target with 3 hbas. The
discovery information provided to the Management platform could see that I
have ONE initiator and a target with say 3 Port Groups.

      With this I can map the initiator to the target, or the initiator to
1 or more port groups.  I would like to the same idea to the initiator name
to include a field like that so that I could easily expand the choices to
maybe map a target port group to an initiator HBA.



               1 --            -- 1

 Initiator(I)  2 --  Network   -- 2  Target(T)

               3 --            -- 3



 I.e. Before I could do the following:

 I -> T

 I -> T1

 I -> T2

 I -> T1, T2

 ETC...



 If I added the info to the initiator name I could also do the following
mappings:

 I1 -> T1, I2 -> T2

 I1 -> T, I2 -> T

 Etc..





 - Stephen