SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


    [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

    Re: iSCSI: Canonical Targets


    • To: ips@ece.cmu.edu
    • Subject: Re: iSCSI: Canonical Targets
    • From: Stephen Bailey <steph@cs.uchicago.edu>
    • Date: Mon, 14 May 2001 16:44:41 -0400
    • In-Reply-To: Message from "Jim Hafner" <hafner@almaden.ibm.com> of "Mon, 14 May 2001 08:10:42 PDT." <OF6DB867BC.B3E95050-ON88256A4C.005293BF@LocalDomain>
    • References: <OF6DB867BC.B3E95050-ON88256A4C.005293BF@LocalDomain>
    • Sender: owner-ips@ece.cmu.edu

    Jim,
    
    > I have a question for you (and the list) about a long Text response (as
    > might be needed for SendTargets).  Can the Text response come in multiple
    > PDUs?  
    
    As I interpret the spec (emphasis on interpret---I don't think the
    spec says anything clearly one way or another), yes.
    
    > In particular, can one "key=value" pair span multiple PDUs?
    
    Not as I interpret the spec.
    
    > And if so, does this help address your resources concerns?
    
    No.  The problem I'm talking about is handling an unbounded TOTAL
    amount of data returned from a single request.  It doesn't matter
    whether it comes in a single PDU or broken into a million, unless you
    can offer a finer grain of flow control for the million PDUs than for
    the one.  In either case the data returned may a) clog the channel b)
    possibly consume resources necessary for other operational purposes.
    
    Again, if we disallow SendTargets (and other interactions with
    indeterminate, potentially large uncontrolled flows) in an operational
    login, we can allow the transport to provide the flow control.  In an
    operational login we don't want to engage transport flow control in
    this way because it interferes with other operational communication.
    
    Steph
    
    


Home

Last updated: Tue Sep 04 01:04:42 2001
6315 messages in chronological order