Bandwidth Measurement for Tor Server Nodes - hibernating the server code only ? ...

[Cav]

Hi or-dev,

I have just spent a little time thinking about this and tinkering here.
I may not have been too clear before.

I ideally want to have my Tor installation acting as a server and a 
client at the same time.
Failing that 2 machines are required, one as a server and one being used 
as a client, which I think is not ideal.
If I could find a way to have it run as a server - a fast one - but with 
limited bandwidth allocated, and then a client at the same time, I would 
be happy.
Right now I have 384mb Recv and 78mb served, so the 2 numbers are not 
necessarily symmetric.

At present with Tor, as you say, EITHER Recv OR Sent is used in 
bandwidth limiting, with connections being blown away when the node 
moves to hibernate state.

Ideally I would like only 'server'' connections blown away, but the 
ability to act as a client intact so I can continue browsing with the 
same node.
In this endeavour I have been playing with the code and so far I have 
half the story covered, with only Sent bytes operating in the hibernate 
code.
What I need now is to switch off the blowing away of 'Client mode' based 
connection, and still enabling them be created. Though I am not sure if 
that's possible ?

Of course, according to how the originators of Tor intended, I could be 
barking up the wrong tree ?

With kind regards,
Cav Edwards



Damian Johnson wrote:
> It hibernates when *either* the read or write total reaches your limit 
> for the accounting period. This is documented in the man page under 
> AccountingMax:
> http://www.torproject.org/tor-manual.html.en
>
> Usually these values are decently symmetric. Cheers! -Damian
>
> On Thu, Aug 12, 2010 at 11:28 AM, Cav <cav at gotadsl.co.uk 
> <mailto:cav at gotadsl.co.uk>> wrote:
>
>     Dear Group,
>
>     I have noticed that if I set a bandwidth restraint for my server,
>     it hibernates when the amount of data 'downloaded' rather than
>     'sent' reaches the specified limit within the accounting period.
>     Thinking about it... does it make more sense for the limit to be
>     applied to the amount of 'sent' data as opposed to 'received' ?
>     -- 
>
>     With kind regards,
>     Cav Edwards
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.torproject.org/pipermail/tor-dev/attachments/20100816/b10fc0c8/attachment.htm>