<html>
<body>
<font size=3>At 09:28 AM 6/7/2005, Fab Tillier wrote:<br>
<blockquote type=cite class=cite cite="">> From: Roland Dreier
[<a href="mailto:roland@topspin.com" eudora="autourl">
mailto:roland@topspin.com</a>]<br>
> Sent: Tuesday, June 07, 2005 8:38 AM<br>
> <br>
> Michael> Why not just use the IETF draft
for RC / UC based IP over<br>
> Michael> IB and not worry about creating
something new?<br>
> <br>
> I think we've come full circle. The original post was a
suggestion on<br>
> how to handle the fact the the connected-mode IPoIB draft requires
a<br>
> network stack to deal with different MTUs for different
destinations<br>
> on the same logical link.<br><br>
That's right - by implementing IP segmentation in the IPoIB driver when
going<br>
over UD, the driver could expose a single MTU to the network stack,
thereby<br>
removing all the issues related to having per-endpoint MTUs.<br><br>
Keeping a 2K MTU for RC mode doesn't really take advantage of IB's
RC<br>
capabilities. I'd probably target 64K as the MTU.</blockquote><br>
The draft should state a minimum for all RC / UC which should be the TCP
MSS. Whether one does SAR over a UD endpoint independent of the
underlying physical MTU can be done but it should not require end-to-end
understanding of the operation, i.e. the send side tells its local that
the TCP MSS is X while the receive side only posts 2-4 KB buffers.
This has been done over Ethernet for years.<br><br>
Mike</font></body>
</html>