Besides the runtime scaling characteristics, I'm very concerned with the scalability of human maintenance. When hosts and services are coming into and going out of service on a regular basis, keeping a monitoring framework upto date with the current state of the inventory can be quite onerous. While Nagios provides a reasonable facility for keeping track of resource availability, it provides no help with keeping the inventory current. After poking around the perldoc for Nagios::Object, it looks like some thought gone into programmatic generation of Nagios configuration stanzas but it seems kinda laughable that Nagios itself can't be programmatically updated; it has to read config files to know what hosts and services to monitor. Ya know what they say: there are high maintenance girlfriends and low maintenance girlfriends and then there are the worst kind... the high maintenance ones who think they are low maintenance. Nagios is a high maintenance girlfriend. Since I don't have it I sure hope someone finds the time on their hands to make the Nagios configuration data-driven; I'd like to have a RDBMS housed inventory system tied into my monitoring framework.
Oh, wait. I worked on such a beast already in an endeavor that begat Hyperic.
( Aug 16 2004, 10:09:25 PM PDT )