[Babel-users] Restarting MeshPoint – seeking advice on routing for crisis/disaster scenarios

[Juliusz Chroboczek]

> Babel (Juliusz, correct me if I am wrong) would most likely drown in new
> iterations of the getting path metric and blackholing a lot of
> routes...

Not quite.  The protocol will keep updating its metric, but then two
things will happen:

  1. the metric oscillations will be dampened by the hysteresis algorithm;
  2. since the router is short on CPU time, it will naturally delay the
     route selection procedure.

In principle, we could artificially delay route selection when there's
a lot of churn, but we haven't found it to be necessary.

> if the input of the routing protocol (link up/down and metric) changes
> faster than your protocol can update its global state, you are in
> trouble (at best).

That's certainly true in a link state protocol.  In a loop-avoiding
protocol (such as EIGRP, Babel, or BGP), you can safely delay route
selection.

Note that some implementations of link-state protocols delay SPF
recomputation under churn (notably implementations of IS-IS), and it seems
to work satisfactorily even though there's no theoretical basis for doing
that.

> Instead of torturing your routing protocol with constantly updating
> the metrics, you need to filter...

...or use a robust hysteresis algorithm.

-- Juliusz