Irqtune: some stats (was Fix for your serial/PPP problems)
I've done some statistics regarding irqtune and kernels 2.0.14 and
2.0.15.
The protocol was the following:
1) ftp to my ISP
2) put a file
3) get the same file 10 times with a 'mget file file <etc...>'
I timed only the 'GET', not the 'PUT' as Linux buffers outgoing
packets and tricks the figures.
Test I was done on a 2.0.14 kernel with a 100kb nul file (dd
if=/dev/zero...)
Test II was done on a 2.0.14 kernel with irqtune and a 100kb nul file.
Test III was done on a 2.0.15 kernel with a 100kb nul file.
Test IV was done on a 2.0.15 kernel with irqtune and a 100kb nul file.
Test V was done on a 2.0.15 kernel with a 100kb random file.
Test VI was done on a 2.0.15 kernel with irqtune and a 100kb random
file.
The same phone line was used for tests I and II, III and IV, and V
and VI.
That is I didn't disconnect between I and II, III and IV and V and VI.
However I rebooted and disconnected between II and III, IV and V.
Here are the timings as reported by ftp, as well as the transfer
rates as reported by ftp.
Then come the average figures for timings and transfer rates. The
average for transfer rates were computed from the average time
(100kb/<avg. time>).
Finnaly, I computed standard deviance for timings (not for transfer
rate, they should be the same as timings if I remember correctly my
stats (arghhh) lessons).
I II III IV V VI
Count Time Rate Time Rate Time Rate Time Rate Time Rate Time Rate
----------------------------------------------------------------------
1 56.4 1.8 27.2 3.7 59.2 1.7 27.3 3.7 67.2 1.5 26.9 3.7
2 68.2 1.5 27.2 3.7 77.0 1.3 36.5 2.7 44.9 2.2 26.9 3.7
3 40.7 2.5 41.0 2.4 58.4 1.7 27.1 3.7 80.0 1.2 26.9 3.7
4 56.7 1.8 37.0 2.7 55.5 1.8 27.0 3.7 47.6 2.1 27.0 3.7
5 63.0 1.6 36.6 2.7 44.8 2.2 41.1 2.4 68.1 1.5 27.0 3.7
6 65.9 1.5 36.5 2.7 58.0 1.7 41.1 2.4 45.6 2.2 26.9 3.7
7 68.9 1.5 41.2 2.4 46.2 2.2 41.0 2.4 62.1 1.6 26.9 3.7
8 55.9 1.8 41.1 2.4 55.4 1.8 41.2 2.4 46.4 2.2 26.9 3.7
9 73.2 1.4 42.2 2.4 70.0 1.4 36.5 2.7 60.4 1.7 27.0 3.7
10 42.8 2.3 36.6 2.7 60.8 1.6 27.2 3.7 48.6 2.1 26.9 3.7
----------------------------------------------------------------------
Avg. 59.17 1.7 36.66 2.7 58.53 1.7 34.60 2.9 57.09 1.7 26.93 3.7
Dev. 10.32 5.18 9.15 6.31 11.58 0.05
The figures are quite clear. Irqtune really improves my serial
performances, even with kernel 2.0.15 (see previous messages, 2.0.15
included a patch which might have helped serial performance).
What I don't understand is why a file consisting of zeros only do not
get the same transfer rate as a random file. I suspect an obscure PPP
problem.
My machine is an AMD DX2-66 with 32MB RAM, 32MB swap, and I have fast
serial controllers (16550A). The modem is an external 33.6 V.34/V.FC
Cardinal connected to my ISP which has 33.6 too.
I'm not subscribed to the linux kernel mailing list, but you can
forward this message over there.
I stick with irqtune right now. BTW, I saw no ill side effect when
using it.
Phil.
Reply to: