[tor-talk] help needed to stress-test an onionbalanced HS - everyone is invited
s7r at sky-ip.org
Fri Jul 3 11:05:18 UTC 2015
-----BEGIN PGP SIGNED MESSAGE-----
OnionBalance doesn't modify the rendezvous spec. Basically it just
uses the HSDir system to publish descriptors signed with the master
key containing introduction points to the failback servers. When you
connect to http://eujuuws2nacz4xw4.onion/ you are in fact connecting
to another hidden service, one of the failback servers. OnionBalance
doesn't modify how Tor builds and handles circuits.
I think there is a controlport command to drop all existing circuits
(similar to New Identity in Torbutton).
On 7/3/2015 12:08 PM, Frédéric CORNU wrote:
> Le 03/07/2015 10:56, Roger Dingledine a écrit :
>> On Thu, Jul 02, 2015 at 11:10:03PM +0200, Frédéric CORNU wrote:
>>> running : watch -n 20 wget -O /dev/null
>> I should point out that this approach should do one rendezvous
>> with th
>> onion service, and then re-use that circuit for each following
>> until the circuit fails at which point it builds one more and
>> While normal Tor circuits expire 10 minutes after the activity
>> circuits to onion services expire after 10 minutes of inactivity.
>> We chose that difference because the rendezvous process is so
>> expensive relative to normal Tor circuits.
>> A better approach here would be to flush the circuits and hidden
>> descriptor between fetches -- the onionperf tools that various
>> folks are working on has code for this I think. But it would
>> indeed be more complex than what you are doing here. :)
> I'm not very familiar with the server-side aspect of HS. Maybe s7r
> can give us details about how OnionBalance works, related to
> incomming circuits. So we can tell which one of the approach you
> describe is best for his tests.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (MingW32)
-----END PGP SIGNATURE-----
More information about the tor-talk