2

Op 02/12/16 om 10:07 schreef Fabio Pietrosanti (naif) - lists:

I'm trying to stress some very small dedicated server with ViaNano and Atoms and would like to try out multiple Tor relay with AES hw acceleration to see the limits

-naif _______________________________________________ tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays

On 2 Dec. 2016, at 20:07, Fabio Pietrosanti (naif) - lists lists@infosecurity.ch wrote:

I'm trying to stress some very small dedicated server with ViaNano and Atoms and would like to try out multiple Tor relay with AES hw acceleration to see the limits

-naif

You can run a minimal local network using chutney, with an increased amount of data, and it will report the bandwidth used:

package-manager install tor git checkout https://gitweb.torproject.org/chutney.git cd chutney # use the tor in your $PATH ./tools/test-network.sh --flavor basic-min --data 1073741824 # use the compiled tor in src/or cd ../tor ../chutney/tools/test-network.sh --flavor basic-min --data 1073741824

You should see output like:

... Verifying data transmission: (retrying for up to 60 seconds) Connecting: Exit to 127.0.0.1:4747 via client localhost:9003 Transmitting Data: .................................................. Single Stream Bandwidth: 63.15 MBytes/s Overall tor Bandwidth: 252.59 MBytes/s ...

The "Single Stream Bandwidth" is the measured bandwidth of the Tor client stream from client through to exit.

The "Overall tor Bandwidth" is that bandwidth multiplied by 4 (client, guard, middle, exit), as an approximation of the bandwidth you might expect from a single tor instance using all your cores.

There are limitations: * the measurement is only approximate (but the more data, the more accurate: I recommend 1GB - 10GB if your machine can handle it), * chutney runs 4 tor processes that each try to use all the cores, so there may be some contention, or, alternately, if you have many cores, there may be some idle cores, * python imposes some overhead sending, receiving, and verifying, * large amounts of data can stall chutney on some machines, * the figures can be hard to interpret, and * the local performance may not match the network performance of a public relay for many reasons.

However, in my experience, the reported transmission speed is within an order of magnitude of the capacity of the server for tor traffic, at least on Linux servers.

For more details, read: https://gitweb.torproject.org/chutney.git/tree/README#n59

T