Polling systems at predefined intervals can be used to gather utilization or
other statistical data from various components of the system and to check
how well services that the system provides are working. The information
gathered through such historical data collection is stored and typically used
to produce graphs of the system’s performance over time or to detect or
isolate a minor problem that occurred in the past. In an environment with
written SLA policies, historical monitoring is the method used to monitor
SLA conformance.
Historical data collection is often introduced at a site because the SAs
wonder whether they need to upgrade a network, add more memory to a
server, or get more CPU power. They might be wondering when they will need
to order more disks for a group that consumes space rapidly or when they
will need to add capacity to the backup system. To answer these questions,
the SAs realize that they need to monitor the systems in question and gather
utilization data over a period of time in order to see the trends and the peaks
in usage. There are many other uses for historical data, such as usage-based
billing, anomaly detection (see Section 11.1.3.7) and presenting data to the
customer base or management (see Chapter 31).
Historical data can consume a lot of disk space. This can be mitigated by
condensing or expiring data. Condensing data means replacing detailed data
with averages. For example, one might collect bandwidth utilization data
for a link every 5 minutes. However, retaining only hourly averages requires
about 90 percent less storage. It is common to store the full detail for the past
week but to reduce down to hourly averages for older data.