Video: Path Discovery in Transparent Bridging and Routing
In the previous video in this series, I described how path discovery works in source routing and virtual circuit environments. I couldn’t squeeze the discussion of hop-by-hop forwarding into the same video (it would make the video way too long); you’ll find it in the next video in the same section.
The video is part of How Networks Really Work webinar and available with Free ipSpace.net Subscription.
Actually, you should have a hierarchy of failure detections since each method has its advantages and disadvantages. --- cross-layer alarms are the fastest but limited to a single link --- BFD is fast, can be multiple hops, but only detects usually 20% packet loss, maybe down to 10%, but not more and cannot detect other QoS issues --- complex active probes with controllers could detect such QoS threshold as 0.5 % packet loss, 15 ms jitter, 100 ms delay. --- integration with end-point measurement and feedback to the lower network layers for path selection
The final solution is simulcast with diversity paths. Then you do not need any failure detection, but it has a price. In some networks, it is cheap, just people are not used it. For radar sensors, or safety critical voice, you normally use simulcast end-to-end. And then you have no convergence problem, no microloop problem.
It still a problem with complex probing, how much overhead you have, what is the precision and accuracy, how many samples you need, if you really share the fate with controlled traffic, if you really measure or just estimate, can you use passive observations, etc.
Failure can mean a totally different thing for each application class and traffic class. Already the ITU-T standards acknowledge this and define failures differently for each class.
So BFD is very useful, but only a small part of the overall story...
@Bela: Thanks a million - insightful as always ;)