DMVPN - Spoke to Spoke Bandwidth / Traffic Tracking

Has anyone out there figured out a good way to pull DMVPN spoke to spoke communications bandwidth as a separate statistic than the normal hub to spoke bandwidth statistic?

I notice on all our DMVPN sites all the data we pull is only the hub to spoke interface so all we see is that entire interface as % busy, but we would really like to be able to see a break down of how busy it is talking to 2 locations.

#1 traffic to the hub site from the spoke

#2 traffic to the spoke site from the spoke

I am not sure if this something we can do inside Orion using SNMP pollers and NHRP (on Cisco routers) or if this is something we need to do using Netflow and maybe IP groupings?

I am open to hearing how others may have resolved this similar issue or they just ignore it and leave this as expected results.

Find more posts tagged with

dmvpn

SNMP

netflow

bandwidth

Accepted answers

All comments

jlintner

My environment has about 700 field sites and two major data centers, all using DMVPN. If you want to monitor interconnectivity traffic ingress/egress for every site with NPM, you should be able to monitor the "tunnel" interfaces but that will only give you overall utilization. Plus, you should adjust your tunnel interface bandwidth in NPM for each device/tunnel so it matches the physical WAN circuit bandwidth.

I think to get more detail than that, you'll need NetFlow Traffic Analyzer but you probably wouldn't want all of your sites sending NTA data to an Orion NTA reciever all the time. I think you would kill it plus eat up some bandwidth. We turn on/off netflow as needed for sites that complain about the network to see what their issues are, which usually turns out to be users wasting time browsing the internet...

mhh351

All,

According to my Cisco Tech Rep, the MIbs for data stats does not exist .....yet.

We are going to go down the BGP pairs road and see if we can device a set of stats that way. It will take some interpretation at the outset, for the IP addresses to be aligned to a device, but we are going to see if we can make sense of it.

One other thing that I should caution, since we are all using DNS for our Node Names and that DNS has the LOOPback0 address in it, that is the only good way to say (IP Address) = (Devicename).

Of course, in my net, it isn't quite working out that way. If you do a BGP Trap to Orion, you will see what I mean.

Anyway, if someone else has a better idea, please share.

Mark