Monitoring Taxonomy

Chapter 2. Monitoring

A Few Types of Monitoring Systems

Throughout the taxonomy, I’ll refer to this or that tool being a “centralized poller” or an “APM” tool, and on, so let’s get some jargon straight before we move ahead.

Centralized Pollers

The oldest tools we describe as “monitoring tools” are centralized pollers. They are usually single hosts that sit somewhere on the network and, like the unfortunate employee in those cell-provider commercials, constantly asks the question “can you hear me now?” to every host you configure it to “monitor.” If any of those hosts fail to respond, the system sends you a notification (typically via SMS, but more commonly these days via an alerting service like PagerDuty or VictorOps).

Strict centralized pollers are notoriously difficult to scale and configure. They also typically work on a one-minute or greater “tick,” which limits their resolution potential in the context of monitoring for performance versus availability.

Passive Collectors

Most modern monitoring systems sit passively on the network and wait for updates from remote agents. Passive collectors always employ agents, which enables them to scale better than centralized pollers. Many systems also toe the line, using both active polling and passive collection as needs require.

Roll-Up Collectors

Roll-up patterns are used often in the high performance computing world where we find thousands of individual on-premises hosts running CPU or memory-bound processes. The idea is to ...