SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


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

    RE: iSCSI: Marker negotiation - draft 11



    --- Michael Schoberg <michael_schoberg@cnt.com> wrote:
    > [(MI +  8) * n - 8] 
    > 
    > This is the formula in the draft.  I'm just
    > wondering how a 2048 byte
    > interval results in a marker at location 1032. 
    > Using the above formula for
    > n = 0 .. X:
    > 
    > n = 0, 		-8
    > n = 1,		512
    > n = 2, 		1032
    > n = 3,		1552
    > ...
    > 
    > That doesn't look like a 2048 byte interval to me. 
    > Perhaps someone could
    > explain the math. 
    
    It clearly says in the 3rd paragraph of A.2: "As an
    example, if the marker interval is 512...". So
    there is no need to mention 2048 when discussing
    this _example_. That's all it is.
    
    Oh, I think I get where the confusion is coming
    from...
    Bytes vs. words, right? Well, I understand that the
    intervals are discussed in bytes, MI is in bytes.
    However, negotiation is in 4-byte words. I.e.,
    sending IFMarkInt = 512 results in MI for T to I
    direction being 2048.
    
    
    > You may want to describe the formula a little
    > clearer:
    > 
    > 	Start of marker (byte location) = ((MI + 8) * n) -
    > 8 
    > 
    > 	[n] is an incrementing integer describing the n'th 
    > 	  marker inserted into the TCP output stream.  n >
    > 0
    > 	[MI] is the negotiated marker interval value in
    > bytes.
    
    I'm afraid it didn't become clearer. I like what's
    in draft 11. The problem with your "definition" of n 
    is that it makes it look like the number of the
    marker,
    which it is not, unless the login consumed MI bytes
    or less.
    
    Adding "in bytes" to MI might clarify things just a
    bit, however, and I don't oppose it.
    
    > Where does this come from?
    > 
    > 	"512+8+512=1032 (-1 since it starts at 0) = 1031" 
    > 
    > What formula are you using?
    
    The same one, n = 2, MI = 512.
    
    
    > (BTW: Why was the above naming chosen over:
    ITMarker, 
    > TIMarker, ITMarkInt, TIMarkInt.  
    > OF & IF prefixes don't seem intuitively
    > perspective-obvious.)
     
    Most everything in iSCSI is defined w.r.t. the
    initiator. "Out" and "In" is used for Data PDUs
    and other things. Now the same philosophy is aplied
    to markers as well. It doesn't much matter what the
    names are as long as they describe the direction
    by themselves. I'm happy.
    
    
    > Instead of offering a range (which seems kind of 
    > unnecessary), how about offering only the value the 
    > entity wishes to receive?  Example:
    
    You may already do that. Offering a range is more
    general. I assume there must have been a reason 
    for introducing ranges. Replacing comma by : or -
    would be nice, however.
    
    > The draft says "Default is 2048", how does that get 
    > negotiated?
    > 
    > I: OFMarker=yes | IFMarker=yes
    > T: OFMarker=yes | IFMarker=yes
    
    Correct, IMHO.
    
     
    > Will the above negotiate the default 
    > OF & IF Marker Intervals?
     
    I think so. 8192 _bytes_.
    
    > Also, section 9.10.4 does not explicitly say that 
    > only a parameter negotiated by the initiator may be 
    > returned in a target's response.  So would this be 
    > possible?
    > 
    > I: OFMarker=yes | IFMarker=yes | OFMarkInt=1,512
    > T: OFMarker=yes | IFMarker=yes | OFMarkInt=1 | 
    >    IFMarkInt=1
    
    Yes, but the way I understand it, it would require
    a followup of:
    
    I: IFMarkInt=1
    T: (nothing)
    
      Martins Krikis, Intel Corp.
    
    
    __________________________________________________
    Do You Yahoo!?
    Try FREE Yahoo! Mail - the world's greatest free email!
    http://mail.yahoo.com/
    


Home

Last updated: Tue Mar 12 19:18:08 2002
9076 messages in chronological order