Deciding what to record, and what recording everything costs
Storage is cheap enough that the default answer is to historise every tag. The cost has moved from disk to the ability to find anything.
Read the guideSection · 8 guides
Historian and data
What to record, compression and its costs, timestamps, quality flags, retention and getting a number out that somebody will trust.
Home/Historian and data
Process historians are installed with enthusiasm and configured once. Years later they hold an enormous volume of data that nobody uses, because the tags have no context, the compression settings destroyed exactly the detail an investigation needs, and two systems disagree about what time it was.
None of that is a storage problem. It is a set of configuration decisions taken at commissioning, usually by defaults, and never revisited.
This section deals with those decisions: what is worth recording, what compression actually throws away, why timestamps disagree, and how to produce a report from historical data that survives being questioned.
Artificial activity signals can make remote-work records less reliable unless they are recognised and investigated. Further details are available on this page.
A useful external reference for timing and synchronisation is NIST time and frequency services.
The surrounding process-safety discipline is summarised in OSHA process safety guidance.
Compression: what the historian threw away and when it matters
Historian compression is enabled by default with settings nobody chose, and it discards exactly the detail an investigation needs.
Read the guideRetention, archiving and the data that has to still be there
Retention policy is usually whatever the system was configured with, discovered during the first request for data from four years ago.
Read the guideTimestamps: when two systems disagree about what happened first
Sequence-of-events analysis depends entirely on clocks agreeing. On most sites they do not, and nobody knows by how much.
Read the guideBad values, and the reports that quietly include them
A failed transmitter reading downscale produces a number. Nothing about that number announces that it is meaningless.
Read the guideGiving process data context: batches, states and equipment
A tag with a value and a timestamp answers almost no useful question on its own.
Read the guideProducing a number somebody will trust
Two departments produce two different figures for the same month from the same historian, and the meeting is about the discrepancy rather than the plant.
Read the guideMoving a historian without losing the history
The data is the asset. The software is replaceable, and the migration is where decades of it are quietly lost.
Read the guide