Three choices...

1. Use the -x and/or -X options to limit the number of hosts ntop processes
to what fits in memory.  Crude, but maybe workable.

2. Use filtering to limit it to the important hosts and/or some of the
workload reduction options - man ntop.  Better than #1, but takes more
knowledge of your environment.

3. Buy more memory.  512MB DDR is still under US$70 if you shop carefully.
Two Saturday's ago BestBuy ad, for example, PC2700 DDR 512MB US$90 less
US$35 Mail-in-rebate.

4. Use a netFlow collector (nProbe, etc.) on the local machine sending the
data to the remote.

5. Use the capture files - but isn't the transport of them causing more
bandwidth usage???

OK, that's 5, but ... you get the drift.


-----Burton


> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] Behalf Of
> pfeito
> Sent: Wednesday, April 28, 2004 7:54 AM
> To: [EMAIL PROTECTED]
> Subject: RE: [Ntop] Post processing of tcpdump files with NTOP
>
>
>
> Hello again,
>
> > -----Original Message-----
> > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On
> Behalf Of Burton
> > M. Strauss III
> > Sent: quarta-feira, 28 de Abril de 2004 4:21
> > To: [EMAIL PROTECTED]
> > Subject: RE: [Ntop] Post processing of tcpdump files with NTOP
> >
> > Which version of ntop?  3.0 is MUCH more stable than the 2.2
> series.  And
> > should have no problems.
>
> Im using 3.0.
>
> >
> > You are wrong about memory usage - ntop doesn't benefit from running
> > off-line, in fact will probably need more memory because it
> won't be able
> > to
> > purge inactive hosts.
>
> I believe so, but in this specific case, the data collecting PC has only
> 256MB Ram which becomes exausted after +- 10H when using NTOP in real-time
> processing mode (due to the large amounts of traffic in the network).
>
> I also have a remote machine with 1GB ram which I can use, but
> not connected
> to the target network, therefore I can only use it to process previous
> collected data. It takes more memory, but with 1GB I could
> process more than
> than 10H.
>
> Ideally, it would be better if I could deploy the 1GB ram machine in the
> target network and use NTOP in real-time, but this is not the case,
> unfortunely.
>
> >
> > Certainly the data you're looking for is in the rrd files - you may need
> > to
> > create some custom graphs using rrdtool, but the data is there.
>
> That is interesting. I've to research further on that. I don't have a clue
> how can I make custom graphs with rrdtool, since I dont know much about
> rrdtool, only that is used to collect periodically.
>
> Thanks for the feedback :)
> -pfeito
>
> >
> > -----Burton
> >
> >
> > > -----Original Message-----
> > > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] Behalf Of
> > > pfeito
> > > Sent: Tuesday, April 27, 2004 7:28 PM
> > > To: [EMAIL PROTECTED]
> > > Subject: [Ntop] Post processing of tcpdump files with NTOP
> > >
> > >
> > > Hi to all!
> > >
> > >
> > >
> > > I'm trying to do some network analysis in a university department
> > network
> > > and I choose to use NTOP to acquire statistic data. 10 hours
> later NTOP
> > > crashed due to lack of memory (only 256MB were available in the NTOP
> > > machine).
> > >
> > >
> > >
> > > I spent a 2 or 3 hours reading some references, trying to
> understand the
> > > memory limitations of NTOP, and, if I understood well, its is kind of
> > > difficult to do a long run analysis (e.g. 1 week or +) with NTOP when
> > > dealing with medium size to large networks, although it really depends
> > on
> > > the machine specs.
> > >
> > >
> > >
> > > I decided to try a different approach: to collect raw tcpdump
> output for
> > a
> > > week, and then feed that data to NTOP. I've done a little
> > > experiment with an
> > > 1 minute tcpdump file and it seem to work well.
> > >
> > >
> > >
> > > Will this method work for 1 week tcpdump file ? I suspect that the
> > memory
> > > limitation still poses a problem, but I could do post processing in
> > > different machine (i.e. with 1GB Ram). It seems to me that
> this offline
> > > processing method should need less memory compared with real-time
> > > processing
> > > mode.
> > >
> > >
> > >
> > > Any feedback from people that has actually done some data processing
> > like
> > > this would be appreciated :)
> > >
> > >
> > >
> > > BTW: as I saw in another post, the tcpdump file only worked when one
> > > specific interface is indicated with -i parameter (e.g.
> tcpdump -i eth0
> > -w
> > > dumpfile)
> > >
> > >
> > >
> > > -pfeito
> > >
> > >
> >
> > _______________________________________________
> > Ntop mailing list
> > [EMAIL PROTECTED]
> > http://listgateway.unipi.it/mailman/listinfo/ntop
>
>
> _______________________________________________
> Ntop mailing list
> [EMAIL PROTECTED]
> http://listgateway.unipi.it/mailman/listinfo/ntop

_______________________________________________
Ntop mailing list
[EMAIL PROTECTED]
http://listgateway.unipi.it/mailman/listinfo/ntop

Reply via email to