On Thu, Oct 18, 2012 at 1:16 AM, Aaron aagbsn@extc.org wrote:
On Sun, Oct 7, 2012 at 5:46 AM, tor-admin tor-admin@torland.me wrote:
Hi,
I am operating two 1 GBit/s servers which host 7 exit nodes. One server is located in GB with
torland1 https://atlas.torproject.org/#details/4E377F91D326552AAE818D5A17BC3EF79639C2...
torland2 https://atlas.torproject.org/#details/332895D092C2524A3CDE8F6E1498FFE665EBFC...
torland1 is configured to run at maximum bandwidth, torland2 runs trottled to ensure not to exceed the 100TB contractual traffic limit. The nodes on this server get traffic as expected and are at the top of the Blutmagie Tor Status http://torstatus.blutmagie.de/index.php?SR=Bandwidth&SO=Desc.
The second server is located in RO and hosts
torland3 https://atlas.torproject.org/#details/E5DB403858511BB166405E4C43776D933A7B86...
torland4 https://atlas.torproject.org/#details/E0D2F8B2A18516A5C935477B06048D7BD2E7EC...
torland5 https://atlas.torproject.org/#details/D2F1745C11843BC5AF09CB9D4027AD2BE5F1B2...
torland6 https://atlas.torproject.org/#details/43F9022028F8AB0F7420F702CB079DF0CA0C80...
The configuration for the GB and RO nodes is:
Nickname TorLandX ORPort 443 DirPort 80 Address a.b.c.d ORListenAddress a.b.c.d DirListenAddress a.b.c.d OutboundBindAddress a.b.c.d
RelayBandwidthRate 30 MBytes # max 30 RelayBandwidthBurst 100 MBytes # max 100
MaxOnionsPending 250
NumCPUs 4
On both servers AES-NI is used by openssl.
Although the server in RO has the same configuration, same operating system and a similar hardware, it does not get traffic as the server in GB. At the moment Atlas shows for torland3-6 only an advertised bandwidth rate of less the 3 MB/s. The GB node torland1 shows more than 30 MB/s.
In order to test if there are any traffic limits I did several speed tests by doing up and downloads of a large 4GB iso image.
Downloading simultaneously with 10 curl instances from different debian mirrors I measured around 1 GBit/s. To measure upload I stored the debian 4GB iso image on the webserver of the RO server and used curl on the GB server to download it from RO. The maximal measured upstream speed on the RO server was around 750 Mbit/s. Therefore I looks for me as there are no traffic limits inplace by my ISP or their upstream.
One of the servers operated by Torservers.net is located in the same data center and runs wau, gorz and sofia at high speed.
I am wondering why the directory authorities only measure such a small fraction of the available bandwidth for the RO nodes. Because I am paying personally for the server I would like to get most out of the server. Can someone who runs a tor authority please help me to understand what is the reason for the low observed bandwidth.
Thanks,
Torland
tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
I am not a directory authority. Have you taken a closer look at the configuration of the Tor exits wau or sofia? They are run by torservers.net -- see this page in particular. https://www.torservers.net/wiki/setup/server#high_bandwidth_tweaks_100_mbps
I might guess that TCP window scaling and latency could be a factor.
--Aaron
Forgot to CC you in my last response. Also see: https://trac.torproject.org/projects/tor/ticket/7177 Perhaps you can add some comments/details to this ticket?
Thanks!
--Aaron