This metric tracks how often an event’s characteristic (timing, duration, or size) falls within a certain range. A distribution is broken up into buckets that represent a specific range of values. Each bucket has a count that increases monotonically. When an event’s characteristic falls within the range for one of the buckets in the distribution, the bucket’s counter is incrementally increased.
A typical example is the page end-to-end time distribution. This distribution is broken into 11 buckets. The first bucket has a range from 0-8 seconds. The second bucket has a range of 8-12 seconds. The last bucket (11th bucket) includes all times greater than 44 seconds. The events of interest are page downloads. Each page download event has the time it took to download the page. For each page event, the download time is used to select the correct bucket from the page end-to-end time distribution. Then the counter for the selected bucket is incrementally increased. If a page event had a download time of 10 seconds, it would cause the second bucket to be incrementally increased.
Percentage metrics indicate what proportion an asset represents as part of the whole. They indicate the levels of usage for assets such as appliance hard disk space and CPU availability. Percentage type metrics can also provide an alternative view of entities that are also represented by count-type metrics, providing frequency or rate of occurrence instead of a total number of occurrences.
This metric tracks how often an object is referenced over a time interval. Reference counters are similar to count metrics but also include a reference to a specific object such as a URL or a web site. Some examples of reference counters are top referrers and exit pages.
Standard metrics provide the mean (or average), minimum, maximum, and standard deviation values observed over a given time interval. They are utilized for measurements that fluctuate up and down over a period of time.
A good example of a standard metric is Page End-To-End Time. Mean page end-to-end time represents the average download time for all pages. Maximum page end-to-end time would be the most lengthy download for the interval, while minimum page end-to-end time would be the shortest download for the interval. Finally, the standard deviation for page end-to-end time indicates how spread out the page end-to-end times are.