Configuring BFD Intervals
In case of BFD implemented in software, the ability to run low-range sub-second intervals depends on the CPU load, which in turn depends on other protocol configurations and system conditions, such as software forwarding. It also depends on switch model.
To find a suitable low-range interval for the chosen configuration and network load, statistics of missed hello packets can help. These hello packets are exchanged between BFD peers, and if n number of packets are missed in a row, path failure will be declared, where n is detection-multiplier. However, missed hellos less than n number will not declare failure, but if statistics are steadily increasing, it can indicate that the currently chosen BFD interval is not appropriate.
In that case, a next longer interval/ higher detection multiplier can be chosen for the BFD session, and statistics can be monitored again. In addition, when statistics are observed periodically, missed hellos can give an indication of whether health of the path is deteriorating. These statistics are collected continuously for each BFD session. The commands to display the statistics of missed hellos are provided in Displaying BFD Information.