Calculating the 'system.mem.used' metric
The manner in which Datadog calculates the system.mem.used metric produces a value that may sometimes be different from what might be displayed by common system resource reporting tools.
For example, running ‘free -m’ on an Ubuntu machine may produce the following memory breakdown (values represent megabytes):
| | | | | |
|---|
| total | used | free | shared | cached | available |
| 128831 | 1203 | 71975 | 4089 | 55653 | 122380 |
A Datadog Agent running on this same machine reports a system.mem.used metric with a value of 56856 MB—clearly different from the ‘free -m’ used memory value of 1203 MB.
The reason for this discrepancy is that Datadog includes cached memory in its formula for used memory, where ‘free -m’ does not.
Datadog calculates used memory as follows:
- system.mem.used(56856) = system.mem.total(128831) - system.mem.free(71975)
Again, Datadog’s system.mem.used metric includes cached memory, so subtracting this cached memory from used memory results in the following value:
- system.mem.used(56856) - system.mem.cached(55653) = 1203
1203 MB—identical to the used memory value reported by ‘free -m’ in the example above.
The system.mem.usable metric represents free memory plus cached memory plus buffers (on Linux, it reflects “MemAvailable” attribute from /proc/meminfo whenever possible).
Additional helpful documentation, links, and articles:
