RE: [opennms-discuss] Software looks great. but I can't run it.

[<jr...@bl...>]

On Tue 6 Feb, 2001, Weston Bustraan <we...@it...> wrote:
>other open source packages. Why not take a cue from packages like MRTG?
>Which are scheduled to run from cron every so often, read their config
>file, poll the required systems, send the data somewhere (generate a
>graph, in the case of MRTG) and shut down. These existing packages don't

MRTG is not a great example. Its polling is very inefficient. Not only
that, it chews up lots of CPU recreating the graphs all the time, even
if no one looks at them. (Cue RRDTool and its front-ends, such as Cricket.)

But even Cricket will take a lot of CPU time if you are polling a lot
of systems: and for efficiency's sake (e.g. host is dead?  don't try and
poll it for lots of different values) it helps to be able to cache some
state about what you're collecting - I'd be inclined to do that in some
form of database, though, and that's presumably where OpenNMS's use
of PostgreSQL comes in.

I find memory and CPU requirements tend to come down to what you're trying
to do - the size of your network, etc - and you've not really said what
that is. 

James.