[tor-relays] A bit more evidence on circuit creation storms

[Gordon Morehouse]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512

This is quite a story - and I've found 0.2.4.x on the Pi to not have
nearly the problems of its predecessor (and .17-rc to be a lot better
than .16-rc)...

Dan Staples:
> Just to add my experiences to the mix:
> 
[snip]
> 
> Finally, I noticed that bandwidth-related config options had no
> effect on the 3 Mb/s traffic flood during the circuit creation
> storms. I had:
> 
> RelayBandwidthRate 200 KB RelayBandwidthBurst 200 KB 
> MaxAdvertisedBandwidth 200KB
> 
> ...yet, still 3 Mb/s traffic floods. Even MaxOnionsPending 250,
> NumCPU 1, and AvoidDiskWrites 1 made no difference in my RPi's
> ability to weather the storms. I eventually had to use QoS on my
> DD-WRT router to set limits on the traffic it would pass to the
> Pi.
> 
> I will try your builds of 0.2.4 to see if that makes a difference.

I guess maybe TCP handshakes could've made up part of that.  If this
is the case, it's a (low priority at the present time, I can imagine)
bug where load on popular hidden services should be distributed to
Guards more effectively.  I dunno if there's a ticket for it or not.

BTW, my builds so far contain nothing special, they are textbook
builds from the Debian source packages but done on a Raspberry Pi so
the armhf is runnable on other Pis.  I will likely start experimenting
with tuning in the future.

Best,
- -Gordon M.

-----BEGIN PGP SIGNATURE-----

iQEcBAEBCgAGBQJSMU7dAAoJED/jpRoe7/ujn+oIAIShPPNyRAOHyri2TuelvnnB
eITzVtngjyz7b+Y2i0kZYqJH5ELWHjjDbd2pRx8QusU52Qb1wv5nxdIYJx1Uj9Q5
vTEOAlNiAZZVKuZPc7p3y1lPmEUppLlNXrYeTFsMDRSiLBejJSgRWjWJsZa1kyqk
MBN4qSvWbTwO9ZVhYCnLuO82KZADNEx1otLSHQ07C/KF3K0MpqM4D6aiAHuQhiv5
hGOGRI6h4JPWuVs2e1xRwot8U1tZY5t19EB+TXRppGUSD6HOA5ZJnFpZRaI7ZUBZ
3stkHuqPCqpUY6BYBvElAhBmiNlX07QUMT1alxA8MlCywsHa9BOlzQiRf9kQhmc=
=obAy
-----END PGP SIGNATURE-----