@Deependra Yadav wrote:
I appreciate your reply and was wondering if you could elaborate further on your thoughts. I understand that while both BFD and IPSLA may collect similar statistics, the way they do so is different. I think that in making final decision by a router to steer traffic based on BFD could be faster than deciding to steer based on PBR relying on IPSLA.
Yes, I agree BFD could be faster. The reason "could", much depends on the frequency of either the SLA or BFD exchange. BFD, should be able to do it faster, as it was designed for lightweight exchanging "liveliness" with a peer.
However, my prior reply's point is there's a difference between detecting "brown outs" vs. "black outs". "Brown outs" are looking for a spike in latency and/or packet loss, something IP SLA was designed to do. "Black outs" are looking for a total lost of a path, something BFD was designed to do (quickly, and with minimal overhead).
I don't know what enhancements were made to basic BFD for AAR (if any). I'm also mostly unknowledgeable about SD-WAN.
I can say, years ago, I worked with OER/PfRv1, which managed performance of flows to all outbound destinations. Under the covers, it used both NetFlow and IP SLA for its analysis. It worked well enough, that our concurrent performance monitoring stopped seeing WAN performance issues, because OER/PfR would route around issues. I suspect AAR's development linage might be based on the early OER/PfR technology.
My guess, either will likely satisfy the basic need, but one of them might be better than the other for specific situations and/or resource consumption on the network devices.
Hopefully, the above provides the additional information you're looking for about my thoughts. If not, and if you can specify a specific concern, let me know and I'll try to clarify my thinking even further.
