ASR 9000 VPNv4 Adding latency

[tham89]

Hi there. 

Using RSP440 / Typhoon LCs / all -TR SKUs (even at the edge where design would call for -SE) we're observing ~20ms latency added for the case where the router is treating traffic entering on a VRF.

MPLS-LDP and MP-BGP provides next-hop traffic decision-making, but the latency is introduced at the first hop, but only under the condition that the traffic entered through the VRF. The latency is not observed in the default VRF over the same exit link.

Looking at CEF in the VRF shows that the flow should be express-forwarded based on the fact that the information in the FIB recurses all the way from the VPNv4 IPv4 destination to the labels used for local-switching, but there is still 20ms latency near ingress.

This is a small deployment and isn't seeing much utilization, where else can I look to explain this anomaly?

[M02@rt37]

Hello @tham89,

You could analyze the path selection and forwarding behavior within the VRF. Check the routing tables, forwarding tables/forwarding decisions for the VRF to understand how the traffic is being processed. Verify that the correct paths are being chosen, and there are no unexpected loops or suboptimal routes that could introduce latency.

Also, check the label switching mechanisms to ensure efficient forwarding within the VRF.

Use packet tracing and monitoring tools to capture and analyze the packets as they enter the VRF. This can help identify any anomalies, delays, or unexpected behaviors at the first hop within the VRF.

[tham89]

These are all covered in my OP, I'm not even talking about a path.

1. I checked the FIB using CEF commands.
2. A traceroute suggests there exists an unexpected delay as the packets egress to the first hop.

[MHM Cisco World]

reduce the MTU by 8 and check again