[tor-talk] hiddden service on openwrt

[coderman]

On 4/27/15, Griffin Boyce <griffin at cryptolab.net> wrote:
> ...
>    I've run hidden services successfully on TP-Link WDR-4300 -- they have
> 8mb of flash (storage), 128mb ram, and 2 usb ports.  With an added flash
> drive it's basically a small server.  For the size-conscious, you can
> also run a hidden service on a TP-Link 3040, but it gets a lot trickier
> as you only have 32mb of ram to work with.

8mb flash, 32mb ram is probably the lower bound for a useful Tor
client / hs host device. remember to omit geoip from the tor pkg :)



> ... For all sorts of reasons I
> wouldn't recommend being a bridge/middle/exit node while running a
> hidden service (especially on a router, which is already extremely
> resource-constrained).

i would state this as: never use these devices as relay.
 and only as bridge if not already running hidden svc,
  and not more than a few distinct onions per device.


a longer hidden service performance discussion spans:

"Hidden Services are in a peculiar situation. While they see a loyal
fan-base, there are no dedicated Tor developers to take care of them."
 - https://blog.torproject.org/blog/hidden-services-need-some-love

"[tor-dev] Hidden Service Scaling"
 - https://lists.torproject.org/pipermail/tor-dev/2013-October/thread.html#5556

"[tor-talk] Hidden Services - how to implement something like Round Robin DNS?"
 - https://lists.torproject.org/pipermail/tor-talk/2014-September/thread.html#34951

"Rumors that hidden services have trouble scaling to 100 concurrent connections"
 - https://bugs.torproject.org/8902
  https://trac.torproject.org/projects/tor/attachment/ticket/8902/torhs-pyloris-nov9.tgz


finally, consider:
Increase NumEntryGuards to 3 or 5
 and Increase MaxClientCircuitsPending 300



>    OpenWRT and Commotion both ...

you'll want to squeeze the best reduction across toolchain (uClibc
opts) and application (for example, ConstrainedSockets opts) even
system tuning (ulimits, /proc). some of this is on the Torouter page:
https://trac.torproject.org/projects/tor/wiki/doc/Torouter the rest on
the list archives...

in general terms, there is not a good set of measurements for Tor
client/hs serv performance. most performance attention focuses on
relays or crypto primitives.

i'd like to run common Tor client/hs-serv benchmarks across a variety
of platforms. in particular the i.MX53 and i.MX6 line as step up
compute wise from these units. e.g.
http://www.inversepath.com/usbarmory.html for a nice i.MX53 device.

best regards,