RRD stuff
my understanding of the RRD database is as such, that it does not store actual points entered into it, rather it stores the values of a function. The database consists of DS's (Data Sources) and RRA's (Round Robin Archives). A DS defines a value to be kept track of. An RRA defines what to do with the values, either AVERAGE, MINIMUM, MAXIMUM or LAST. All rrd's here have RRA's defined for AVERAGE. That is how ganglia creates a database for each type of metric that it monitors. The relationship of a DS to RRA is one-to-many. A single DS can be compiled into multiple RRA's. A metric is usually averaged.
The process, as I see it is as follows:
some process (ganglia's gmetad) updates the database with a PDP (primary data point). The database performs a function (average) on the PDP, creating a CDP (consolidated data point) which actually gets stored.
The process, as I see it is as follows:
some process (ganglia's gmetad) updates the database with a PDP (primary data point). The database performs a function (average) on the PDP, creating a CDP (consolidated data point) which actually gets stored.