init.d script for multi instance TOR servers

[Mike Perry]

Thus spake Moritz Bartl (moritz at torservers.net):

> On 06.09.2010 01:25, Mike Perry wrote:
> > Please keep us posted on your total capacity (using a network monitor
> > like munin, mrtg, or the console-based nload). I find that without killing
> > irqbalance and manually setting the smp_affinity, I can't push more
> > than 400Mbit without a CPU hitting 100% utilization (run top then hit
> > '1' - look at the idle column). Moritz has found the same. 
> 
> That sounds like you are able to push more than that after tweaking? My
> CPU usage is okay now (around 50-70%), but I am still unable to push
> more than an average of 200 Mbit/s per direction, with a maximum of 240
> Mbit/s. Judging from the Torstatus page, yours don't perform much better?
>
> http://us1.torservers.net/stats/graph_6.html (M = MByte/s)

I have pushed 450Mbit per direction max so far, and seem to be
averaging about 350Mbit now according to nload. I just added some more
node processes the other night to see what that changes.

> Only Blutmagie seems to be able to push more, which suggests that
> - there is a bottleneck in Tor
>
> - there is indeed not more demand for Tor traffic on the ports we allow

These two seem unlikely. Simply running more nodes should attract more
traffic. Which is more or less what I've experienced, but I need more
time to determine if the effect is linear.

> - there aren't enough middle nodes?

Hrmm.. According to the bw authorities (which anyone can observe by
checking out torflow.git and running the statsplitter.py using a Tor
client):

0-100%: N: 1884, Bw: 2976.18, X: 772, XBw: 1500.11, BT: 21, Dirs:894, Up: 18.9
Overall Cnt: 1884
Overall Avg Ratio: 1.19916910039
Overall Ratios < 1: 67.7%
Overall Ratios < 1 Avg: 0.494694530226
Overall Ratios > 1: 32.3%
Overall Ratios > 1 Avg: 2.67405428424

Guard Cnt: 423
Guard Avg Ratio: 1.69746927915
Guard Ratios < 1: 35.0%
Guard Ratios < 1 Avg: 0.642766635274
Guard Ratios > 1: 65.0%
Guard Ratios > 1 Avg: 2.26509106567

Mid Cnt: 689
Mid Avg Ratio: 0.969512524368
Mid Ratios < 1: 80.3%
Mid Ratios < 1 Avg: 0.435115123618
Mid Ratios > 1: 19.7%
Mid Ratios > 1 Avg: 3.14246666124

Exit Cnt: 629
Exit Avg Ratio: 0.844718499259
Exit Ratios < 1: 83.9%
Exit Ratios < 1 Avg: 0.503680814752
Exit Ratios > 1: 16.1%
Exit Ratios > 1 Avg: 2.62756896876

Guard+Exit Cnt: 143
Guard+Exit Avg Ratio: 2.39079031284
Guard+Exit Ratios < 1: 32.2%
Guard+Exit Ratios < 1 Avg: 0.631389792777
Guard+Exit Ratios > 1: 67.8%
Guard+Exit Ratios > 1 Avg: 3.22514519864


It does look like middle nodes and pure exits seem to be the most
overloaded, but if you also look at the 10% breakdowns, they are not
off by much in any particular slice.


-- 
Mike Perry
Mad Computer Scientist
fscked.org evil labs
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://lists.torproject.org/pipermail/tor-relays/attachments/20100905/8f117ede/attachment.pgp>