[Cu-wireless] Information System

[Wendy Edwards]

> If you go through the archives of the list you'll see that last fall we
> discussed a way to forward logs back to a central server using SMTP.
> That could be used to provide historical network adjacency information
> (how many times did a link flap, etc.) without the requirement for
> additional processes on the routers.

At the risk of asking a dumb question, would that mean that the nodes
would have to have SMTP services running?  Would there be any risk of
spammers exploiting our network?

> Regarding real-time information, the routers will provide routing
> updates on all their interfaces (at least in the current state of the
> project, using OSPF). So that could be used to directly obtain network
> topology information passively (=without having to query a router). In
> turns, reachability information can be obtained & collected using
> regular tools -- fping, mrtg.

We could probably use that to collect routing, topology, and reachability
information.  What traffic data?  Do we want to know how heavy
the load is for a node or group of nodes?  What if two nodes are having a
lot of collisions?

> I guess I'm just saying there are ways to collect the necessary
> information without having to heavily instrument the routers (embedded
> systems: Soekris = 486 CPU + 64MB RAM). Just an idea...

Yeah, I think it is really important that this is designed to minimize the
burden on the nodes.  (That was one of the things that motivated the idea
of an aggregate server.)