Re: [lwip-users] Which opt.h options effect TCP receive throughput?

[Kieran Mansley]

On Tue, 2008-01-22 at 14:09 -0500, Bill Auerbach wrote:

> 
> #define TCP_WND                         (64*1024-1)

Sounds OK.  You could perhaps get an wireshark capture to see if your
windows are becoming empty to let you know if this is big enough.

> #define TCP_MSS                         (8*1024)

Too large.  The MSS size on your system (as you say you're not doing
jumbos, is much more likely to be 1500 bytes) or thereabouts.  This
won't harm performance but may be using up memory that could be used for
other things.


> One thing I notice is there is an ACK sent for every 2 packets
> received.  Does this sound right?  (2874 received, 1437 sent).

That's correct.

> What is the theoretical limit with TCP?  Is it the PC slowing things
> down, or am I simply close to “as good as it gets”?

There are I suppose a couple of theoretical limits: 
 - your physical layer bandwidth minus the TCP/IP header overhead
(likely to be a few percent)

 - if you are window limited: bandwidth = TCP_WND / round trip time.
i.e. if you can increase the TCP_WND or decrease the round trip time you
should get increased bandwidth.

You say that "98% of the processor is processing lwIP timers and
checking for Ethernet packets".  That seems like a lot, and if your CPU
is saturated like that you're probably at the limit of your performance
without offloading something like checksum calculation if you aren't
already.

Kieran