Re: A question on Zero Copy

[Jeff Chase <chase@cs.duke.edu>]

Option A requires only that the NIC support payload recognition or
payload alignment.  I like the term "payload steering" for option B, and
it is supported by the RDMA proposals.

It seems to me that iSCSI needs payload steering if it is trying to look
like a block device to an off-the-shelf Unix operating system.  For
example:

1) If you build an iSCSI card that plugs into an SCSI slot and talks
iSCSI on a network link, the card needs payload steering.

2) If you build a block device driver that talks iSCSI through a network
card, the card needs payload steering.

Otherwise you have to copy the payloads.

Jeff


"Randall R. Stewart" wrote:
> 
> Hello all:
> 
> Has I am trying to get a grasp of this issue, maybe
> someone can fill me in on a conflict I have in
> my resolution processing :)
> 
> Does the iSCSI layer want:
> 
> A) Plain Zero copy, where the upper layer (iSCSI) asks
>    to read the next available "message" from the wire
>    into a buffer passed to the transport by iSCSI?
> 
> <OR>
> 
> B) A directed Zero Copy, where the upper layer (iSCSI) asks
>    to read a particular request to a specific buffer?
> 
> (A) being what I thought as normal zero copy semantics whereas (B)
> is something quite different... it implies I can issue several
> read requests passing buffers to the underlying transport and
> wanting specific things to go in each buffer...
> 
> For some reason I may have incorrectly gotten the impression
> that iSCSI wants (B), which as I think about it is a
> hard problem, or am I just confused and iSCSI wants to use (A)????
> 
> Thanks
> 
> R
> --
> Randall R. Stewart
> randall@stewart.chicago.il.us or rrs@cisco.com
> 815-342-5222 (cell) 815-477-2127 (work)