On Fri, 26 Jan 2007 14:40:48 +0100, Stephane Bortzmeyer <[EMAIL PROTECTED]> 
said:

> On Fri, Jan 26, 2007 at 02:12:53PM +0100,
>  Alexander Gall <[EMAIL PROTECTED]> wrote 
>  a message of 1149 lines which said:

>> We happen to have full NetFlow data for this period of time, but
>> haven't analyzed it yet.  It would be interesting to see, for
>> example, how the distribution of queries per IP adddress

> IMHO, this is the first thing to do. The rate, if I read correctly
> your graph, is very low (40 kb/s) and one single broken machine could
> generate it.

I should have mentioned that we see the same pattern from all other
peers and our transit providers.  Also, I wouldn't expect to see the
typical daily variations in this case.  The total traffic to this
server is a few hundred kilobits per second which translates into a
couple hundred queries per second (most queries are below 100 bytes).

I should really look at the flows, but that isn't trivial because they
are stored on tape together with about 3 years worth of flows of our
network.  I'll have to talk to the researchers at ETH Zurich who
analyze this data for various projects.

--
Alex


_______________________________________________
DNSOP mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/dnsop

Reply via email to