Statistics Types

AMPS collects statistics in several different ways. For each individual statistic, understanding how the number is collected and calculated -- that is, which type of statistic you are working with -- is important for accurately understanding the statistic.

The table below lists the statistics types in AMPS:

Type	Description
Cumulative	An accumulated count or total for the current instance or monitored object lifetime (for example, bytes received on a client connection since it connected).
Recovered Cumulative	An accumulated count or total that is recovered from the latest persisted statistics sample when AMPS restarts and continues accumulating. With a new statistics database, it starts from zero. Examples include total_samples, total_time, and total_commit_time in the statistics resource.
Fixed	Information that is fixed for the lifetime of the instance or monitored object (for example, the process ID of the AMPS server).
Snapshot	The value of a metric at a specific point in time (for example, the number of active subscriptions for a client when statistics are collected).
Interval Snapshot	The number or amount of events recorded during the preceding statistics interval, without dividing by the interval duration (for example, bytes received by a host network interface since its previous sample). A newly observed host network interface establishes a baseline and reports zero for its first interval snapshot.
Interval Average	An average computed over the statistics interval (for example, bytes sent per second on a host network interface, or the percentage of CPU time used during the interval).
Running Average	An average computed over the lifetime of the monitored object (for example, the conflation ratio for a conflated topic). When an average uses recovered cumulative totals, it can span instance restarts, as described for the individual metric.