OrdinateDB

Migration

SPEC 11 §4

Bring your history with you.
All of it, at your own pace.

Moving a plant’s history is a considered decision, so the path is built to be gradual, reversible and provable at every step. Your current historian keeps running throughout; nothing is deleted, and nothing is taken on faith. Here is the migration path we actually run.

The migration path

SPEC 11 §4 · SPEC 04 §6

Five steps. None of them a leap of faith.

The whole path is designed so that at every stage the old system still works, the new system is being proven, and pausing is cheap. You keep full control of the pace from first extraction to cutover.

Many sites use the migration moment to stand up a unified namespace at the same time. The two moves compose: the broker carries current state, the historian keeps the memory — OrdinateDB and the UNS.

The tooling is open source and the procedure is a spec, not a statement of work — read the PI engineer’s guide.

  1. 01

    Extraction collector

    extract

    Reads the PI archive through supported PI APIs only — nothing your support contract objects to. Resumable and checkpointed, so a week-long extraction survives a restart at hour 90. Throttled, so the plant keeps running on PI, untouched, while its history is read out.

    supported APIs · checkpointed · rate-limited

  2. 02

    AF XML import

    model

    Your Asset Framework structure — elements, attributes, templates — imported from AF XML into the OrdinateDB model. Duplicates that accumulate in any long-lived estate are surfaced for a decision instead of being copied forward, so the model arrives cleaner than it left.

    AF XML → templates + assets · dedupe gates

  3. 03

    Parallel run

    parallel

    Dual collection into both systems while trust builds. Both historians see the same live data from the same sources, for as many weeks or months as your confidence requires. You cut over on evidence, at a moment you choose.

    dual collection · no cutover pressure

  4. 04

    Validation

    validate

    The compat: pi query mode reproduces PI’s retrieval semantics — boundary handling, interpolation, summary behaviour — so historical queries can be compared digit for digit against the system your reports were built on. Confidence you can show your quality manager.

    compat: pi · digit-matching comparison

  5. 05

    Cutover

    cutover

    Collection points switch when you say so, and not before. The PI archive is retained read-only for as long as you like — a migration is not required to end with a deletion.

    PI archive retained read-only

What moves, and what gets rebuilt

SPEC 11 §4

Your data, structure and history move as they are. A few things come out better than they went in.

01

Displays are rebuilt, better

Model-driven displays are generated from templates: one template per asset class, not one hand-drawn display per asset. Most sites come out of a migration with fewer, more consistent displays than they went in with — and every one of them versioned.

02

Calculations are reimplemented, clearer

Legacy calculation code is reimplemented as OrdinateDB expressions or WASM modules. The results are typically shorter, easier to read, and — often for the first time — documented and version-controlled. It is also a natural moment to review what each calculation should be doing.

03

Interfaces are assessed case by case

Most custom interfaces map directly to existing OrdinateDB collectors; a few justify a bespoke one. We will say which is which before any work is scoped — and occasionally an interface turns out to serve equipment that retired years ago, which is worth knowing too.

All three are scoped at the start of a migration plan, not discovered at the end — which is what makes the plan dependable.

Two ways to run this

Do it yourself, or benefit from our experience.
The tooling is free either way.

Route A

Guided migration

Our consultancy has run this path before, and it shows in the plan. A fixed-scope, fixed-price assessment inventories your tags, AF structure, displays, calculations and interfaces, and returns a written migration plan you can act on — with us or on your own.

Migration assessment
Route B

Do it yourself

The extraction collector, AF importer and validation tooling are free, Apache-2.0 and documented to the same standard as the product. Plenty of sites migrate on their own — the docs are written for exactly that.

Read the migration docs