On my backbone I have some PEs with 1072 running only OSPF + MPLS + iBGP to be able to run VPLS. In just 1072 I have very high processing, almost 50% of the processing consumed by MPLS verified in tools> profile. In other PEs, my processing is all below 6%.
In PEs, I don't run any package inspection, so I get the fast path. I use in ip> rpflter setings for DDos.
This high processing PE is running ROS 6.46.8 (long-term) and firmware 6.46.8.
The routing table has 508 routes.
I have already exchanged for another 1072 to see if there was a problem with a cash bug and even then the problem continues.
Has anyone caught a similar problem?