Link availability detection
Continuous link detection#
The principle behind the availability detection of a Link is the tranmission (and receiving) of a continuous stream of packets from one TAG to another.
Each TAG monitors each TAP by sending packets through the TAP to the other end.
As long as data is being received, the TAP is considered active and is used to pass traffic.
The default convergence time for the suspension of use of a TAP from a Tunnel Aggregation Group is 300ms. This is configurable in the following ways.
Packets are sent every interval, and a set number of continuous failures triggers the TAP to be suspended.
The default interval is 100ms meaning a total of 3 failures, meaning 300ms convergence time is standard.
The return of a link only takes a single successful receiving packet. In practice this means up to 100ms of time since the return of the link to a usable state.
Configurable per Link#
Some types of circuit (in particular wireless services such as 3G/4G/5G) often benefit from a longer convergence time. This can be related to the charges for bandwidth use. Although small, the traffic per link is continuous, and where charges are applied for usage, the default interval can be lengthened to save bandwidth usage.
Where a finer granularity is required, and the latency of the link allows, the interval can be taken lower than 100ms. This is not recommended for links that are not already very low latency.
The mechanism employed by link availability detection is a stream of gratuitous ARP packets. These are small, adhere to common network standards, and are used for declaring availability of a network service.
Currently there is no automatic flap dampening. TAPs can be disabled manually.
This is being considered for a future release, along with other link quality metric tracking and triggers.