Monitoring with agents or not?

Over the past decade – or more – a discussion has raged regarding the relative merits of basing your monitoring of applications, servers or networks on agents installed on the monitored system, versus monitoring without such agents. Some of the most obvious differences (or advantages) are:

• Agents can collect data even when the connection to the management system is down
• Agents may perform actions independently of the management system
• Agents see things from the inside, agent-less sees them from the outside
• Agent-less requires less software installation
• Agent-less is less intrusive

A lot of parameters may be discovered/monitored irrespective of whether agents are used or not. And for some tools, you may have a free choice. For other tools there are differences in licensing or pricing, favoring one approach over the other. There are frequently also differences in the resolution possible – at least in a simple manner. There may also be differences in security level achieved, e.g. if one approach sends encrypted data over the network while the other does not.

Several agent-less approaches are not really that: They just use a pre-installed agent from somebody else, e.g. the vendor of the hardware and/or software in question. This is true of most SNMP-based monitoring, as well as monitoring that uses remote access protocols such as WMI (Windows Management Interface).