Selection guide
Where to install Historian alongside Citect SCADA
Short answer
Put the Wonderware Historian on a separate server sized for its history files and leave the connector on the Citect SCADA host. The connector only pushes data out, adding no Historian client functions to Citect, and events need their own acceptance test.
On this page
- Should the Historian go on the same server as Citect SCADA?
- What will operators see in Citect once the connector works?
- Where should the connector itself run?
- How much disk does the Historian server need?
- Which Historian features does the connector leave out?
- Do alarms, events and operator logs reach the Historian too?
- How do you prove the feed works before you accept it?
- Questions asked
Put the Wonderware Historian on its own server and keep the Citect-to-Historian connector on the Citect SCADA host. That is the layout to choose for a Citect SCADA 2016 system being extended with history: the SCADA server keeps its CPU, memory and disk for acquisition and operator displays, the Historian gets a machine sized for its history files, and the connector — the Plant SCADA Connector in current documentation — stays beside the tags it reads. Separating the two is not about co-existence, which is not the problem; it is about keeping storage writes and report queries off the host that renders alarms. This article gives the placement decision, the disk number behind it, the connector limits that shape what you get, and the acceptance tests that prove the feed.
Should the Historian go on the same server as Citect SCADA? #
No — put the Historian on its own server and keep only the connector on the Citect host, so history storage and report queries never compete with the services that keep operator screens and alarms responsive. The products co-exist without known conflicts, and small plants do run SCADA, MES and Historian on one box, so co-existence is not the question. Capacity is: a successful installation and quiet-hours operation prove nothing about the same host during peak production while a report scans the history files. That matters beyond convenience, because the host you would be starving is the one rendering the alarms an operator acts on.
There are three layouts worth considering, and the choice comes down to what shares what:
| Layout | What shares what | When to choose it |
|---|---|---|
| Citect, connector and Historian on one server | CPU, memory, disk, maintenance window and failure exposure | Only a small installation with measured spare capacity and acceptable shared downtime |
| Citect and connector on the SCADA host; Historian on its own server | Storage and queries leave the SCADA host; collection still stops with the Citect host | The default for any system large enough to justify a historian |
| Connector on its own host; Historian on its own server | Collection survives a Citect host failure; the collector process itself is still single | When Citect runs redundant primary and secondary servers, or measured connector load hurts the SCADA host |
For the small-system case, decide it this way: a plant small enough that one server comfortably carries both rarely needs a historian at all, and when a compliance requirement forces one anyway, co-host only with measured peak-load headroom and give the history files their own disk volume. And if the real need is a few hundred tags landing in SQL for reports, the decision belongs upstream with the SCADA package itself, as covered in Selecting SCADA software for mixed PLCs and SQL logging.
What will operators see in Citect once the connector works? #
Nothing. The connector pushes data out of Citect into the Historian; it does not add trend retrieval, tag browsing or any other Historian client function to the Citect operator interface. A running feed and a usable historical view are two separate deliverables: the connector delivers the first, and the second needs a Historian client or a separately engineered retrieval interface running where engineers will actually use it. Decide that viewing path before commissioning, or the project delivers a store full of data nobody on the plant floor can read.
Where should the connector itself run? #
On the Citect server by default. Move it to its own host only when your Citect project runs redundant servers, because the connector's primary and secondary server settings then keep collection alive through a Citect host failure. Two facts shape that decision. First, the redundancy protects the data source, not the collector: the connector remains a single process wherever it runs, so its state belongs on a monitoring list, not in a redundancy argument. Second, the connector's constraint on larger systems is collection throughput, not host resources, so a dedicated host buys resilience, not speed — validate throughput against the intended tag population and update rates rather than trusting a small pilot set. When the connector does get its own machine, the role calls for a compact headless box rather than another server: an industrial mini PC mounted with the controller covers that spot.
How much disk does the Historian server need? #
Plan on gigabytes over the retention period, not the 200 MB install figure: the documentation's 200 MB is the free space the history files need at installation, while production size is tags times rate times retention. The AVEVA System Platform 2023 documentation sets that 200 MB minimum for the historical data files to install — a floor for setup, not a sizing basis. Production storage is:
- tags × collection rate × retention × bytes per stored sample
Worked example, with illustrative inputs you replace with your own: 500 analog tags collected at a 5-second rate and retained 2 years (730 days).
- Samples per tag per day: 86,400 s ÷ 5 s = 17,280
- Samples per day, all tags: 500 × 17,280 = 8,640,000
- Samples over 730 days: 8,640,000 × 730 = 6,307,200,000
- At an assumed 12 bytes per stored sample: 6,307,200,000 × 12 B = 75,686,400,000 B ≈ 76 GB
For that duty cycle, give the history files a 256 GB volume: two years fit with better than 3× headroom. Verify the 12-byte assumption against the store's measured growth after the first week of running, because nothing on the collection side thins the store — swinging-door deadband, minimum storage rate and rollover are not connector-controlled options, so it expands at the full configured rate for as long as retention requires. Watch free space as a routine check; the number that decides when to add disk is measured growth times remaining retention, never the install minimum.
Which Historian features does the connector leave out? #
Three storage options and the equipment hierarchy: swinging-door deadband, minimum storage rate and rollover are not available as connector-controlled collection options, and the Citect equipment model does not appear as a hierarchy in the Historian.
| Missing or changed | Effect on the project | What to do |
|---|---|---|
| No Historian client functions in Citect | The feed is one-way; retrieval is separate | Specify the viewing path as its own deliverable |
| Discrete tags stored under a .Discrete destination name | Queries miss the tag if they use the Citect name | Query the created name; keep a mapping register |
| String tags stored under a .String destination name | Same failure mode for string history | Include string tags in the commissioning set |
| Deadband, minimum storage rate and rollover not connector-controlled | The store grows at the full collection rate | Size disk by rate × retention, as in the worked example |
| No hierarchical equipment representation | Destination tags arrive flat | Design tag naming for navigation and reporting |
| Limited connector diagnostics | Faults localize poorly from logs alone | Use controlled tag changes and end-to-end comparisons |
Keep a mapping register holding each source tag, its data type, its created Historian name and its intended use, and include representative analog, discrete and string tags in the commissioning set — inspect the created names before building reports on them. If the project depends on any capability in the table, resolve that requirement before deploying the integration, because a working value feed does not satisfy a storage-policy or information-model requirement. These capabilities are stated for the connector as currently documented; the release notes for your exact version decide whether any of them changed.
Do alarms, events and operator logs reach the Historian too? #
Not automatically. Process values and event records travel as separate paths through this integration, and not every Citect log message reaches the Historian's alarms and events, so historizing operator actions may need added Citect logging configuration. Trends working proves the value path and nothing about the event path. Before relying on this integration for record retention, list the classes you must keep — alarm transitions, acknowledgments, operator commands, system messages — and test each one explicitly. If a required class fails its test, plan the added logging configuration before you commit, and keep any existing logging or historian running until the replacement passes.
How do you prove the feed works before you accept it? #
By testing each boundary separately: prove the live tag first, then the connector mapping, then the stored record, then the client query — a moving value on a Citect screen proves only the first of those four. Work the symptom down the chain:
- A stale or missing live value is an acquisition problem — controller, driver or Citect tag — the class of fault covered in Getting Citect SCADA talking to a Mitsubishi iQ-R PLC; fix it before touching history.
- A live value that moves while history stays absent is a mapping or ingestion problem: confirm the tag is selected for collection, then query the Historian directly for its created name, suffixes included.
- History that exists but shows nowhere in Citect is expected with this integration; retrieval lives in the separate client.
- Trends that work while events are missing mean the event path failed its test, not that the feed failed.
- History that lags the live values points at throughput or host load: compare timestamps across the path and watch both hosts during the same interval.
The decision in one line: Historian on its own server with disk sized by tags, rate and retention; the connector on the Citect host unless Citect itself is redundant, and then on a compact host of its own; retrieval and event capture engineered and tested as separate deliverables.
Questions asked
Can Citect operators trend Historian data once the connector is running?
No. The connection is one-way: it pushes Citect values into the Historian and returns nothing to the Citect operator interface. Retrieval needs a separate client or a separately engineered interface on the machine engineers will actually use, so treat viewing as its own deliverable with its own acceptance test rather than an automatic result of configuring the feed.
Is the documented 200 MB of free disk space enough for production?
No. That figure is the minimum free space the history files need at installation, not an operating size. Production storage follows tags times rate times retention: in the worked example, 500 tags at a 5-second rate for two years came to about 76 GB at an assumed 12 bytes per stored sample. Measure the store's growth over a week, multiply by 52, and add headroom for retention beyond two years.
We already run SCADA, MES and Historian on one server and everything installed fine. Does that validate co-hosting?
It validates co-existence, not capacity. A clean installation and quiet-hours operation say nothing about behavior during peak production while a report query scans the history files on the same disk subsystem. If you must co-host, measure the SCADA server during its busiest period with a heavy historical query running, and accept that maintenance windows and host failures are shared.
Will alarm acknowledgments and operator actions be historized through the same setup?
Treat them as a separate acceptance item. Process values and event records take different paths through this integration, and not every Citect log message reaches the Historian's alarms and events; capturing every operator action can require added Citect logging configuration. List the record classes you must retain, test each one explicitly, and keep any existing logging or historian running until those tests pass.