Re: iSCSI minimum PDU length

[Paul Koning <ni1d@arrl.net>]

>>>>> "Stephen" == Stephen Bailey <steph@cs.uchicago.edu> writes:

 >> The minimum MSS is under 300 bytes.
 Stephen> ...
 Stephen> What you really seem to be talking about here is the minimum
 Stephen> allowable maximum PDU length, NOT, as the subject indicates,
 Stephen> and JohnH points out, the true `minimum PDU length'.

 Stephen> One number relevant to this discussion is the one below
 Stephen> which you would not expect (or desire) good performance
 Stephen> anyway.  When MSS shrinks below this performance limit you
 Stephen> may have to fall back to a slow path (e.g. reassembly &
 Stephen> copying instead of on-the-fly direct placement) in your
 Stephen> implementation, but you won't feel bad because you know
 Stephen> there's no way performance could be good at that MSS under
 Stephen> ANY circumstances.

 Stephen> The minimum allowable maximum PDU length should be NO LARGER
 Stephen> than this number.

 Stephen> I personally think that you will want to maintain
 Stephen> performance at an EMSS smaller than 1024.  I agree with Dave
 Stephen> that 512 is certainly headed in the right direction.

Nice analysis.

I don't see any need to maintain performance at MSS values that
small.  MTU values below 1480 are ancient history at this point, and
sensible MSS values are surely not way below MTU.

So, given the design center of what we're building, 1024 is the
logical value.

The one argument I can see for picking 512 is that it is the value
that was in the spec before.

     paul