Re: Low capped TCP network speeds over the Internet

[mlelstv%serpens.de@localhost (Michael van Elst)]

z411%omaera.org@localhost (z411) writes:

>Best case:
>[ ID] Interval           Transfer     Bitrate
>[  5]   0.00-1.00   sec  0.00 Bytes  0.00 bits/sec
>[  5]   1.00-2.00   sec  1.00 MBytes  8.39 Mbits/sec
>[  5]   2.00-3.00   sec  1.12 MBytes  9.44 Mbits/sec
>[  5]   3.00-4.00   sec   256 KBytes  2.10 Mbits/sec
>[  5]   4.00-5.00   sec  1.88 MBytes  15.7 Mbits/sec
>[  5]   5.00-6.00   sec  1.62 MBytes  13.6 Mbits/sec
>[  5]   6.00-7.00   sec   512 KBytes  4.19 Mbits/sec
>[  5]   7.00-8.00   sec   256 KBytes  2.10 Mbits/sec
>[  5]   8.00-9.00   sec   640 KBytes  5.24 Mbits/sec
>[  5]   9.00-10.00  sec  1.12 MBytes  9.43 Mbits/sec

>So with a increased the value it can occasionally hit higher speeds,
>it's not much. And what's worse, it makes it very unstable.


The "unstability" is mostly packet loss together with slow
recovery.


The slow recovery comes from limiting packet bursts as requested
in early RFCs. Other BSDs have relaxed that a long time ago,
and I'm currently running NetBSD with similar modifications in
the TCP stack.

What I do not yet understand is what causes the packet loss.

To some degree this seems to be a consequence of running the
network stack under KERNEL_LOCK, but that can only be part
of the truth.

Another issue is e.g. the wm(4) driver that easily sends
packets out of order (and that I work around with a small
patch).


With my local modifications, I can now do near Gigabit wirespeed
over a simulated WAN distance of 100ms (RTT). But the results
over the real internet are much worse.