tcp performance

[Michael Vince]

Zongsheng Zhang wrote:

>Hi, *,
>
>For testing throughput of a TCP connection, the following topology is used:
>     Host-A ---GB Ethernet--- Dummynet ---GB Ethernet--- Host-B
>
>Host-A/B use FreeBSD v6.0. Sysctl parameters of Host-A/B are:
>kern.ipc.nmbclusters=32768
>net.inet.tcp.inflight.enable=0
>net.inet.tcp.sendspace=2097152  # 2M
>net.inet.tcp.recvspace=2097152  # 2M
>
>When RTT in Dummynet is set to 0ms, the throughput (A--B) is about
>900Mbps. The buffer size is enough to fill a link bandwidth=800Mbps, and
>RTT=20ms. However, if RTT is set to 20ms, the throughput is only about
>500Mbps.
>
>Are there other parameters which are necessary to adjust? Does anyone
>have suggestion for high throughput?
>
>Thanks in advance.
>
>  
>
What are you following on your hosts? Release/stable or current? I only 
use just release or stable.

For your middle router try net.inet.ip.fastforwarding or net.isr.direct 
but not at the same time, then try on top enabling polling.
Personally I found enabling polling worked best combined with 
net.inet.ip.fastforwarding.

Andre Oppermann claimed in a post just recently he gets best performance 
using just net.inet.ip.fastforwarding without polling but that might be 
for just -current, I am not sure.

You could also try using current, but if I had to guess you already are?

Regards,
Mike