Hi,
yes I know that packets are biggest than 4 bytes on wire, I'm telling
about payload only.
[40(eth)+20(ip)+20(tcp)+4(payload)]*1000(pktps)*8(bits)*500(clients)=336Mbps1Gbps
My intention is to stress the software and understand what happens in
critical situations. If I use bigger packets I can
Do you think that routing cicle is ok? There is a better way to do that?
Can you give any advice about that?
How do you solve the issue which can happen if a client is not receiving
data fast enough ? The sending socket will buffer data (It is asynchronous
on send as well) and all memory could
Hi,
I'm doing some tests to evaluate the performance of my application server.
I'm using a tcp socket server on a single thread (dedicated, I don't use
the main thread), connected with 500 clients installed on different
machines (no clients on the server machine).
What I want to evaluate is the
For this I configured each client to send about 1000 packets per second of
4 bytes (about 32Kbps).
Your bps computation is wrong. You send 4 data bytes, the actual packets on
the network are much larger (something like 10 times) ! Have a lokk there:
http://sd.wareonearth.com/~phil/net/overhead/
