[c-nsp] Mechanisms in gathering Traffic Matrix in Service Provider networks

[Arun Kumar]

Hi All,

We are planning to gather traffic matrix in our network. Our network
comprises of Cisco devices (NPE-G1, NPE-G2, GSR, CRS, 7600, ISR-G1/2).
Currently there is no traffic matrix calculation per service/COS that has
been done in our network for capacity planning. We would like to start an
exercise of gathering traffic matrix to account for POP to POP, PE to PE
traffic for better visibility into our network.

Have gone through a NANOG presentation which explains about traffic matrix
calculation "http://www.nanog.org/meetings/nanog35/presentations/telkamp.pdf
".

I tried doing MPLS traffic counters but that is not giving the correct
value. BGP Policy Accounting is another option but I dont think it will
work for MPLS VPN customers as majority of our traffic is through MPLS VPN.

Our network do not have point to point TE tunnels built between POP to POP
to gather this  information.

I am exploring the option of "MPLS Aware Netflow" in "P" routers to gather
this information.

I would like to hear from others who have done such kind of traffic matrix
in their network along with its pros and cons  in the interim which would
help me in effectively working out?


thanks in advance
Arun