Re: iSCSI: data and data sequences for Read

["Julian Satran" <Julian_Satran@il.ibm.com>]

Paul,

Nothing limits the burst rate except the TCP window.

The bursts for read accomplish several things:

they might be good "turnaround" points for bi-directional commands (minor)
they delimit data that can be (on request) positive ack-ed enabling 
resources to be released at target


Julo




Paul Koning <ni1d@arrl.net>
Sent by: owner-ips@ece.cmu.edu
14-11-01 00:22
Please respond to Paul Koning

 
        To:     ips@ece.cmu.edu
        cc: 
        Subject:        iSCSI: data and data sequences for Read

 

I hope this isn't a dumb question, but between the -08 spec and the
archives I'm puzzled about the details around data sequences in the
case of Read operations.

As far as I can tell, a read can result in one or more data sequences
coming back.  These are numbered with DataSN values that  keep going
up across sequence boundaries.  Each sequence is limited by
MaxBurstSize, but the total Read size (sum of all the sequences) is
not bounded other than by SCSI.

In the Write case, something analogous happens but there there's an
R2T to control the flow of data sequences. 

As far as I can see, there is nothing analogous in the Read case.  In
other words, while MaxBurstSize limits the size of a data sequence,
there is no mechanism limiting the number of bursts, or the rate at
which you send back DatIn PDUs (other than TCP window control).

Is that right or did I miss something?  If it's right, what is the
purpose of having the notion of a Data Sequence for DataIn, and what
does MaxBurstSize do for you in that case?

Thanks,
                 paul