Components of a Statistic
Can the value of a statistic be subdivided into useful parts such that the sum of the values of the parts accounts for one hundred percent of the value of the statistic? Define a partition for such a statistic. If the sum of the values of the parts does not equal the value of the unpartitioned statistic, do not create a partition for the statistic.
For example, //:class.cpu//:stat.usage
shows the total usage of all CPUs in the system, and //:class.cpu//:stat.usage//:part.mode
shows the total idle
, inter
, kernel
, stolen
, and user
usage of all CPUs in the system. Similarly, //:class.cpu//:res.id/0//:stat.usage
shows the total usage of CPU 0, and //:class.cpu//:res.id/0//:stat.usage//:part.mode
shows the total idle
, inter
, kernel
, stolen
, and user
usage of CPU 0. The value of //:class.cpu//:res.id/0//:stat.usage
is equal to the sum of the values of all parts of the //:class.cpu//:res.id/0//:stat.usage//:part.mode
partition.
Partitions of a statistic should have approximately equal capacity. If the capacities vary greatly, consider providing a normalized statistic. For example, users should be able to easily see that a 1GB network card is 90% utilized and a 10GB network card is 10% utilized.