Hello,

We need to transfer data between two embedded systems at a high rate
(~400 Mbit/second sustainable). This must be done over a Gigabit
Ethernet link, fiber or CAT5, not very long (few meters).

Supposing that the transmitter won't do anything except transmitting
the data, and the receiver won't do anything except receiving the data
and writing it down (to a very fast storage), is it feasible to reach
a sustainable 400 Mbit/s stream on TCP/IP ?
On a more general note, wow can I estimate the maximal throughput over
TCP/IP ? And over UDP/IP ? (supposing that the channel is relatively
clean and the transmitter and receiver are on a direct link - no
collisions).

Thanks in advance
Eli

eliben <eliben@gmail.com> wrote:
> Supposing that the transmitter won't do anything except transmitting
> the data, and the receiver won't do anything except receiving the
> data and writing it down (to a very fast storage), is it feasible to
> reach a sustainable 400 Mbit/s stream on TCP/IP ?

Why not go ahead and try it? Either via FTP to include the filesystem
part or with something like netperf:

http://www.netperf.org/

> On a more general note, wow can I estimate the maximal throughput
> over TCP/IP ? And over UDP/IP ? (supposing that the channel is
> relatively clean and the transmitter and receiver are on a direct
> link - no collisions).

A small matter of doing sums. Consider what UDP or TCP do to get data
from one end to the other. Then consider what IP does to get the
datagrams carrying the UDP datagrams or TCP segments from one end to
the other, then consider what the link-layer protocol does to get the
frame(s) carrying the IP datagrams from one hop to the next.

rick jones
--
Wisdom Teeth are impacted, people are affected by the effects of events.
these opinions are mine, all mine; HP might not want them anyway...
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...

On Sep 4, 1:17 pm, eliben <eli...@gmail.com> wrote:
> Hello,
>
> We need to transfer data between two embedded systems at a high rate
> (~400 Mbit/second sustainable). This must be done over a Gigabit
> Ethernet link, fiber or CAT5, not very long (few meters).
>
> Supposing that the transmitter won't do anything except transmitting
> the data, and the receiver won't do anything except receiving the data
> and writing it down (to a very fast storage), is it feasible to reach
> a sustainable 400 Mbit/s stream on TCP/IP ?
> On a more general note, wow can I estimate the maximal throughput over
> TCP/IP ? And over UDP/IP ? (supposing that the channel is relatively
> clean and the transmitter and receiver are on a direct link - no
> collisions).
>
> Thanks in advance
> Eli

You will probably have to tune the TCP window size of machine. If you
are using Dr. TCP might be an easy way to tune the TCP stacks on the
two machines.

--
VisualEther 1.0 - http://www.EventHelix.com/VisualEther
Generate sequence diagram from Wireshark (Ethereal)