[H-GEN] Traffic analysis recommendations, Extended Question

David Jericho david.jericho at aarnet.edu.au
Tue Apr 19 22:12:21 EDT 2005


Josh Marshall wrote:
> My first suggestion would be to use iptables. It can count the number of 
> packets and bytes that match the parameters, and if you don't specify a 
> target the packets pass right past it untouched.

Oh, good Lord no.

This very quickly becomes unmanagable as the method does not scale. The 
complexity and headspace required to manage a ruleset that could 
potentially run into the hundreds or thousands of rules is horrible. 
Don't mix your traffic accounting rules with your firewalling rules.[0] 
Only leads to mistakes being made.

Anthony (De Crow) mentions he'd like to break it down into PIPE vs 
non-PIPE traffic, without some form of automation based on a BGP or OSPF 
feed, could very quickly become worthless.[1] If the user wants this to 
be correct, they should contact their ISP to see if there is some form 
of looking glass.

David de Groot's suggestions of Cacti[2] and ntop[3] are good. Depending 
on the config of the router, fprobe[4] and nnfc[5] maybe handy tools 
too. I know of NFA[6] being used around the traps.

[0] Just as you don't mix heavy spirits and sparkling white - it only 
leads to bad hangovers.
[1] Some admins can't manage bogon lists, draw your own conclusions.
[2] http://www.cacti.net/
[3] http://www.ntop.org/ntop.html
[4] http://sourceforge.net/projects/fprobe/
[5] http://sourceforge.net/projects/nnfc/
[6] http://www.parnet.edu.au/NetFlow.html

-- 
David Jericho
Systems Administrator, AARNet
Phone:     +61 7 3864 8379
Mobile:    +61 4 2302 7185





More information about the General mailing list