Compression Tuning

Overview

This chapter shows how to measure whether PI Points store too many or too few archive events, and how to review and apply the recommended exception and compression settings.

Compression Tuning needs the Admin role, and the PI Data Archive must allow compression tuning writes (see PI Data Archives in the Administration Guide). It is under Tools in the navigation.

Concepts

• A tuning run observes selected PI Points for a while. During observation PI Nexus+ temporarily lowers only the exception deviation, so more source values reach the PI Snapshot. Archive compression stays as it is.
• After observation the original exception settings are restored. PI Nexus+ then compares the detailed Snapshot stream with what the Archive kept and simulates candidate settings.
• Each point gets a classification, such as Undercompressed (stores more events than needed) or Overcompressed risk (may hide real movement), and, where safe, a recommendation.
• Data Fidelity estimates how closely the compressed series follows the reference signal. 97 % suits most process points. It is not sensor accuracy.
• Nothing is applied automatically. You apply, skip or exclude each point.

PI Nexus+ cannot create source events that do not exist: a point that updates once an hour gives one sample an hour, and may end as Insufficient data.

Create a tuning run

1. On Runs, choose Create tuning run and enter an optional Run name.
2. Choose the Scope: a PI Point Rule, or Selected PI Points and add them with Add PI Points.
3. Choose Adaptive samples (Recommended) to let each point finish on its own, or Fixed window to observe every point for the same time.
4. For adaptive runs, choose the Sample target (Quick, Recommended or High confidence) and the minimum and maximum duration.
5. Choose the Tuning profile: Conservative, Balanced or Fast. Lower load means more observation waves; read the estimate and split a large scope if the archive is busy.
6. Tick the confirmation that settings will be adjusted temporarily and restored, and choose Queue.

The run continues in the background; follow it on Runs. You can also start a run for one point from its dialog: More actions > Tune compression.

Review and apply a compression recommendation

1. On Runs, click a run that is Ready. Stay on Ready.
2. Click Apply to list the points with a safe recommendation, or Investigate for those that need a closer look.
3. To act on some points, tick their rows: Selected rows applies only to them.
4. Current filter acts on every point matching the filter, including rows not loaded yet. Check the count before choosing Apply, Skip or Exclude.
5. To look at one point first, open its row menu and choose Compare trend.

1. Show or hide Snapshot, Archive, Recommended and Custom in the legend.
2. Use Pan, Box zoom and Reset to inspect peaks, steps and noise.
3. Compare Configured, Recommended and Custom below the chart, then choose Apply recommended, Skip or Exclude. Switch on Custom to try your own values or target fidelity; the apply button on the Custom panel writes them.
4. Use Previous and Next to move through the points of the list.

Skip marks the point done for this run and changes nothing. Exclude keeps it out of future runs until it is allowed again on Point Exclusions; its settings stay as they are. Every applied change is recorded with the previous and new values in Change History.

A point with a guardrail warning, or an investigation classification such as Zero/span issue, can only be applied from the comparison dialog, one point at a time, with Apply tested candidate.

Tune large populations with Auto Tune

On Auto Tune, create a plan that processes a rule's points, or a fixed selection, in small recurring batches. A plan covers each point once unless you enable retuning with a cooldown. Auto Tune only creates recommendations; you still review and apply them on Recommendations or in the run.

Reference: Compression Tuning Reference