CPU, Memory, Disk, Network, and Availability are the metric categories, but this article covers how iPM actually handles the underlying data — how it's collected, stored, and displayed — which is useful context for reading any graph in the system correctly.
Collection interval
Every service check runs on its own interval (commonly every few minutes, configurable per check). Shorter intervals give finer-grained data and faster alert detection, at the cost of more load on both the monitored device and iPM itself — critical infrastructure often justifies a shorter interval than a low-priority host.
Graph timespans
Every performance graph lets you switch the time window it covers — typically ranging from 1 hour up to a year. Shorter timespans show raw, high-resolution data; longer ones necessarily show aggregated data, since a year of minute-by-minute readings would be unreadable (and expensive to store) at full resolution.
Aggregation: Min, Avg, and Max
For longer timespans, iPM aggregates multiple raw data points into one, and lets you choose how: Min, Avg, or Max. Each tells a different story about the same period:
- Avg smooths out noise and is best for spotting overall trends
- Max highlights the worst moment in each interval, which is often what you actually care about for a metric like CPU or latency, where brief spikes matter
- Min is most useful for a metric where the low point is the concern, such as available disk space or free memory
Reading graphs correctly
Because of aggregation, a long-range graph can visually understate short, sharp spikes that a shorter-range view would show clearly. If a Max-aggregated view over a month looks smooth, that's a genuine signal the metric has been stable; if it isn't smooth, drop to a shorter timespan to see the actual raw behavior behind it before drawing conclusions.
Data retention
Historical data doesn't stay at full resolution forever — most monitoring platforms, iPM included, roll older data up into coarser aggregates over time to manage storage growth. If you need exact raw figures from further back than your environment's retention policy covers, capture or export them (see Reports) closer to when they're generated.