Runtime Metrics

[Michael Wehrmeister]

Hi,

Is there a reference for how to read the runtime metrics after a monextract all? I want to look at CPU utilization, but the .csv files are a bit challenging to understand. For example, after a netperf experiment, I see cpu_system is 16.1 for my netclient vm, but I am not sure what this number is referring to, or what cpu_system is referring to.

Thank you,
-Michael Wehrmeister

[Michael R. Hines]

Hi,

I'd recommend starting with the GUI --- it organizes the metrics in graphical form and allows you to track them during runtime over time.

We do not have explicit wiki-style documentation on the individual metrics, unfortunately.

As a quick summary as organized in the GUI:

All of the VM-specific metrics come from ganglia. So, refer to ganglia-documentation on things like cpu_system, etc.

All of the application-specific metrics come directly from the script that executes the workload in question --- they're are written specifically by the netperf scripts, in this case.

All of the management-specific metrics come directly from the tool and are pretty self-explanatory when you run the GUI. They time all of the individual operations when interacting
with the cloud itself, such as provisioning time, boot time, ssh time, time to application ready, etc.

Of course, we're happy to answer specific questions here on the mailing list.

/*
* Michael R. Hines
* Senior Engineer, DigitalOcean.
*/