Hardware and Software Requirements
What Flow 7.2.0 needs from Windows, SQL Server, and the machine it runs on.
Infohub is a Windows and SQL Server application. Its appetite is driven almost entirely by two things: how many measures the model contains, and how often they need calculating.
Software prerequisites#
| Requirement | Detail |
|---|---|
| Operating system | 64-bit Windows Server, or 64-bit Windows for evaluation and desktop use |
| Runtime | .NET 8 — Flow 7.2.0 components target net8.0, and Flow Config targets net8.0-windows |
| Database | Microsoft SQL Server |
| Installer | 64-bit MSI, per-machine install, digitally signed |
| Browser | Any current Chromium-based browser, Firefox, or Safari for the web client |
The installer is per-machine and requires administrative rights: it registers the Flow Bootstrap Windows service, creates an event log source, and writes to %ProgramFiles% and %ProgramData%.
SQL Server edition. Express is fine for evaluation and small models. For production, use a Standard or better edition — not because Infohub requires a feature Express lacks, but because Express caps database size, memory, and CPU, and a growing Infohub database will meet those caps.
Sizing the Infohub server#
The Data Engine's work is proportional to the number of measures × the number of intervals it must produce, plus the cost of the source queries. These are practical starting points, not hard limits.
| Model size | Measures | vCPU | RAM | Notes |
|---|---|---|---|---|
| Evaluation | under 500 | 2 | 8 GB | Infohub and SQL on one machine is fine |
| Small site | 500 – 5,000 | 4 | 16 GB | Separate SQL Server preferred |
| Medium site | 5,000 – 25,000 | 8 | 32 GB | Separate SQL Server; watch source query time |
| Large site | 25,000 – 100,000+ | 8 – 16 | 32 – 64 GB | Consider splitting collection and web serving across Platforms |
Two things matter more than the raw counts.
Minutely measures are expensive. A measure at the Minute interval produces sixty times as many values as an hourly one, and everything above it in the chain inherits that volume. A model with a few thousand minutely measures behaves like a much larger model. Use minutely only where you genuinely need the resolution.
Source latency dominates. If a historian takes four seconds to answer a query, no amount of CPU on the Infohub server helps. When collection is slow, measure the source first.
Sizing the SQL Server#
Give SQL Server the resources. It is doing most of the real work.
| Model size | vCPU | RAM | Storage |
|---|---|---|---|
| Small | 4 | 16 GB | SSD, 100 GB+ |
| Medium | 8 | 32 GB | SSD, 250 GB+ |
| Large | 8 – 16 | 64 GB+ | SSD or NVMe, 500 GB+, data and log on separate volumes |
Storage growth is driven by interval count and retention. A rough estimate: total annual rows ≈ Σ (measures at each interval × intervals per year). One hourly measure is about 8,800 values a year; one minutely measure is about 526,000. Multiply by your model and add index overhead.
Use SSD or better. Infohub's write pattern is many small inserts and updates across the calculation window, which spinning disks handle poorly.
Ports#
| Port | Used by | Direction |
|---|---|---|
| 4500, 4501 | Flow Bootstrap — component discovery and inter-component communication | Inbound to the Infohub server; between servers in multi-Platform deployments |
| 1433 (default) | SQL Server | Infohub server → SQL Server |
| 80 / 443 | Flow Server web client | Inbound from users' browsers |
Data source connections use whatever port their target system requires — a historian, an OPC server, an MQTT broker, or an HTTPS API. Those are outbound from whichever Platform hosts the connection, which is worth remembering when the connection lives on a remote Platform inside a plant network.
Service accounts and permissions#
The Flow Bootstrap service is created to run as NT AUTHORITY\SYSTEM by default.
Change it to a dedicated domain service account when:
- Infohub must reach SQL Server using Windows integrated security
- A data source authenticates with Windows credentials, as OPC and several historians do
- A consumer writes to a network file share
That account needs to log on as a service, hold appropriate rights on the Infohub database, and hold whatever rights the data sources and consumers require.
Integrated security follows the service account, not you. A connection you test successfully in Flow Config runs as you. The same connection, once deployed, runs as the Bootstrap service account. A connection that works in Config and fails once deployed is very often this difference.
Virtualisation and cloud#
Infohub runs on virtual machines and in cloud IaaS without special consideration. Two cautions:
- Do not overcommit CPU on the host. The Data Engine's work arrives in bursts at interval boundaries, and a starved vCPU at the top of the hour shows up as late data.
- Keep the Infohub server and SQL Server close to each other in network terms. The chatter between them is frequent and latency-sensitive; an Infohub server in one region talking to a database in another will disappoint.
Antivirus and backup#
Exclude Infohub's program and data directories from real-time scanning:
%ProgramFiles%\Flow Software\Flow
%ProgramData%\Flow Software\Flow
Back up the SQL database — that is the whole system. See Backup and Recovery.
Checking what you have#
Before installing, confirm:
- Windows is 64-bit and patched.
- .NET 8 is present, or allow the installer to bring it.
- You can reach SQL Server from the Infohub server and create a database on it.
- You have local administrator rights on the Infohub server.
- You know which account the Bootstrap service should run as.
- You have your licence details to hand — see Licensing Infohub.