A baseline is a moving average value of a monitored attribute. Baseline values are unique to your environment and reflect the performance characteristics of your system.
Using the collected data, the baselines are calculated for numeric type attributes such as counters and gauges. You can use these baselines as references for identifying problems.
Baselines are sliding averages of hourly values. A sliding average means that there is a fixed size window in which to store values. When a new value is added to the window, the oldest value is deleted; therefore, your average is always calculated on the last n values, where n is the fixed window size.
All baselines are averages of hourly summary values. The hourly summary values that are averaged depend on the baseline type.
The baseline types are as follows:
Averages the hourly summary value for one specific hour of one specific day of the week. The maximum window size is 10 (for example, Friday 17:00 for the last 10 weeks).
This is the most granular baseline because you are comparing only the same specific time of the week.
There are 168 hour of day baselines, one for each hour of each day name (24 x 7).
Averages the hourly summary value for all hours in a specific day. The maximum window size is 10 (for example, every Tuesday for the last 10 weeks).
Day of week is not as precise as the hour of day baseline. Every hour on every Tuesday is averaged together, not taking into account the fact that some hours, for example, working hours, frequently have different workloads than others, for example, late shift or off-peak hours.
There are seven day of week baselines, one for each day of the week.
Averages the hourly summary value for all hours in the day for each of the past 30 days. The window size is 30.
Daily is the least precise baseline. Like day of week, it does not account for the workloads of different hours. Also, it disregards the different daily workload patterns, for example, week days are often busier than weekend days, and so on.
There is one daily baseline.