User`s guide

5967–9446 485
Using Baselines
Baselines combine historical measurements with statistical algorithms
to analyze network data.
In determining baselines for your network, NetMetrix gathers individual
measurements over long periods of time and presents this analysis in a
graphical format. The baseline graph shows patterns of activity, trends
in behavior, and exceptional events.
When a single week is viewed, patterns become apparent on a day-to-day
or even a time-of-day basis; longer views show trends that can help plan
for future expansion or reveal subtle problems that are getting worse
over time.
Measurements
Baselines are based on the 1800-second study and are statistically valid
after a minimum of two weeks. In practice, you may need to collect statis-
tics for one or two months to see trends and patterns for your network in
the baseline graphs.
To ensure sufficient data for baseline calculations, configure data collec-
tion for the agent. Refer to the Agent Manager chapter in Data Collector
Reference.
Methodology
Sophisticated data reduction algorithms permit long-term history func-
tions including baselines.
Network measurements are collected into data sets; statistical analysis
is performed for each data set individually, and the results are used to
determine the baselines. The statistics calculated include the mean and
standard deviation and are used to determine low baseline, cumulative
average, and high baseline for each data set.
Two time periods affect how network measurements are assigned to a
data set: a baseline interval and a baseline width. The baseline interval
specifies how many consecutive 1800-second measurements are treated