Re: [pmacct-discussion] Geographical aggregation

2012-11-21 Thread Jérôme Fleury
Hi Frédéric, that is an interesting question. There is no easy way that I know to do what you want as far as I know. Obviously, this advocates the need for a "plugin architecture" inside pmacct, which I strongly support. Also, let me imagine something: - have a special BGP daemon that would be p

Re: [pmacct-discussion] Geographical aggregation

2012-11-21 Thread Jérôme Fleury
Yo, on est juste à côté dis donc :) Quand est-ce que je passe vous voir ? On Tue, Nov 20, 2012 at 8:30 PM, Frédéric GABUT-DELORAINE < frederic.ga...@dailymotion.com> wrote: > Hello everybody, > > First of all, pmacct is a great tool, thanks for the amazing work :-) > > We use it to collect netf

Re: [pmacct-discussion] Geographical aggregation

2012-11-21 Thread Paolo Lucente
Hi Frederic, Jerome, It's correct there is no support for geo aggregation within pmacct and i find it an interesting topic. Generalizing, whatever approach is taken some sort of mapping is required, ie. route to location, router (ie. sFlow exporter, NetFlow exporter, BGP next-hop) to location or B

Re: [pmacct-discussion] Geographical aggregation

2012-11-21 Thread Frédéric GABUT-DELORAINE
Le 21 nov. 2012 à 10:24, Jérôme Fleury a écrit : > > Hi Frédéric, Hello Jérôme, > that is an interesting question. There is no easy way that I know to do what > you want as far as I know. Obviously, this advocates the need for a "plugin > architecture" inside pmacct, which I strongly support.

Re: [pmacct-discussion] ip_dst are all 0s

2012-11-21 Thread Paolo Lucente
Hi, Great to know it worked. Getting to the accuracy part: one suggestion you received what to enable renormalization - did it not work for you? Are you using sampled NetFlow? The alternative issue could be with the filter itself: if you repeat the big file transfer and try with/without the filter

Re: [pmacct-discussion] Geographical aggregation

2012-11-21 Thread Frédéric GABUT-DELORAINE
Le 21 nov. 2012 à 18:54, Paolo Lucente a écrit : > Hi Frederic, Jerome, Hi Paolo, > I see two cases essentially: 1) Router to location and BGP community to > location is easy to solve as the source for that will always be internal > data/docs: a map, with syntax similar to pre_tag_map would do

[pmacct-discussion] Aggregation of sflow and netflow

2012-11-21 Thread Frédéric GABUT-DELORAINE
Hello list, I have a functional nfacctd daemon which collects IPFIX statistics and a BGP full view from our routers. We have old foundry routers which only send sflow :(. I haven't find a way to aggregate both flows into the same memory plugin and use the same BGP lookup. What I want to do is

Re: [pmacct-discussion] Geographical aggregation

2012-11-21 Thread Paolo Lucente
Hi Frederic, On Wed, Nov 21, 2012 at 06:10:54PM +, Fr?d?ric GABUT-DELORAINE wrote: > I agree, our communities defines a lot of geographical information from where > the route was learned, however the aggregation through interfaces id and > router is enough for us. The mapping "country" -> c

Re: [pmacct-discussion] Aggregation of sflow and netflow

2012-11-21 Thread Paolo Lucente
Hi Frederic, Yes, to collect both NetFlow/IPFIX data and sFlow ou will have to start nfacctd and sfaccd separatedly - each with the own BGP daemon. Typical issue here is that routers do not support BGP peering over a port different than 179/TCP: if both nfacctd and sfacctd run on the same server t

Re: [pmacct-discussion] Aggregation of sflow and netflow

2012-11-21 Thread Frédéric GABUT-DELORAINE
Le 21 nov. 2012 à 19:46, Paolo Lucente a écrit : > Hi Frederic, Hi Paolo, > Yes, to collect both NetFlow/IPFIX data and sFlow ou will have to > start nfacctd and sfaccd separatedly - each with the own BGP daemon. > Typical issue here is that routers do not support BGP peering over a > port diffe