Re: New List: rdma@cisco.com: to discuss RDMA

[Matt Wakeley <matt_wakeley@agilent.com>]

Stephen Bailey wrote:

> > However, I would support structuring the spec so that an RDMA
> > transport mechanism could be used underneath (I guess that's
> > motherhood).
>
> Not necessarily.
>
> You have to ask the implementors (particularly the hardware
> implementors), what sort of optional RDMA proposal they'd be happy
> with.  My answer is none.  It's mandatory or not at all.

great! get rid of it!

>
> The reason for using RDMA is to make the implementation of iSCSI
> easier in hardware.  If there are implementations which do not support
> the RDMA protocol, then your hardware implementation will have to
> support both the `easy path' (using RDMA) and the `hard path' (no
> RDMA).  If you have to implement the hard path anyway, there's no
> point in implementing the easy path.
>
> The argument that you could make the hard path infrequent and
> implement it in software doesn't wash in this case.  It just takes one
> implementation that doesn't do RDMA to slow your system to a crawl,
> and the competitor who only implemented the non-RDMA path makes you
> look like a fool.
>
> Fundamentally, RDMA has to be either adopted or punted.  Of course,
> I'm happy to have somebody prove this statement wrong.

Ok, iSCSI does not need an RDMA mechanism because (from David Black's
message):

> > Actually, RDMA is not needed in FCP because all protocol chips
> > implemented perform a real peer-to-peer DMA straight to the
> > data areas specified by the user's interaction with the operating
> > systems allocation algorithms.  The combination of the FCP/SCSI
> > pointer structure, task tagging, and the FC relative offset perform the
> > function you would otherwise have to use RDMA to accomplish.
>
> And this illuminates the design tradeoff that may motivate RDMA.  If
> one only wants to accelerate one protocol (SCSI/FCP in the above
> example) then having hardware understand its headers and doing
> the DMA on that basis is a fairly obvious way to go - HBAs for both
> parallel SCSI and Fibre Channel (SCSI/FCP) do this.

The issue is that TCP does not provide message framing such that the out of
order messages could be found and processed.  So, instead of focusing on
"RDMA", let's focus on "iSCSI message framing".

-Matt

>
>
> Steph