RE: Requirements specification

[Michael Krause <krause@cup.hp.com>]

At 03:09 PM 8/7/00 -0700, psarkar@almaden.ibm.com wrote:



>Mike,
>
>I think we need some convergence here.
>
>There is no denial that there are applications that can take
>advantage of multiple connections. However, none of them
>are particularly simple (they go way beyond round-robin and
>a lot require router assistance as you say) and the point

I do not believe they are complex nor that there is a lot of router 
assistance, if any, required - they do require the ability to take input 
from other policy making entities who have this knowledge and then execute 
a policy algorithm.  This policy may be opaquely communicated via a request 
to establish a new connection on a given path or to issue operations using 
a particular weight round-robin algorithm.  All of this is outside of the 
iSCSI specification and all that is needed within iSCSI is the mechanism to 
communicate the results of these external policy decisions.

I believe there is a functional layering (really lateral than vertical) but 
that this does not require multiple layers within iSCSI itself - the policy 
management, etc. can be implemented outside of iSCSI (a peer functional 
component) that is completely optional for an implementation (it would use 
a default of 1 connection per session).  Those that want to take advantage 
of these capabilities would be able to do so without necessarily having to 
rework their implementation other than to take action input through this 
defined interface.

Mike