Re: we seem to have another huge traffic reduction

[Martin Mulazzani <e0225055@xxxxxxxxxxxxxxxxxxxx>]

Hi!

In the past few days there was a drop in the number of servers with the
guard flag set. See the attatched picture. Number of running, running
exit and running fast servers is as usual.

If you are interested I can check what countries changed the number of
guard servers compared to regular activity.

Kind regards, Martin Mulazzani


Roger Dingledine wrote:
> On Tue, Oct 07, 2008 at 07:42:19PM -0500, Scott Bennett wrote:
>   
>>      Over the last few days, the torstatus pages appear to have confirmed
>> what I see when I watch the activity of my tor server:  another precipitous
>> drop in traffic load over the entire tor network.  This is the second such
>> reduction in total traffic that we've seen in the past few weeks.  Are these
>> reductions due to some of the changes in the last few versions of tor?  Or
>> is something actually wrong in our network?
>>     
>
> Interesting questions. So long as Tor seems to keep working ok for
> most people as a client, I don't think there's something wrong with the
> network. Some of it could be reduced directory overhead -- but directory
> overhead on the server side should be easy to spot in graphs, since it
> should show up as asymmetric bandwidth use.
>
> As I mentioned earlier, we're working on plans to start gathering more
> comprehensive stats on the Tor network, how Tor relays are behaving,
> how client-side performance is looking, etc. This should help us be
> able to answer questions like this in the future. We're keeping some
> data already, so hopefully we can look into the past once we have better
> tools and look for trends. Expect to hear more here in early 2009.
>
>   
>>      I notice that 0.2.1.6-alpha fetches fresh consensus documents far less
>> frequently than 0.2.1.4-alpha and earlier did, typically only once about every
>> three hours instead of roughly hourly.
>>     
>
> There shouldn't be any change between those versions, as far as I know.
>
> It is true that 0.1.2.x fetched a networkstatus document every 30 minutes,
> and 0.2.0.x and later fetch one every 2-3 hours.
>
> --Roger
>
>