We are evaluating Ignition SCADA for an industrial automation application where approximately 100,000 OPC UA tags will be integrated from the control system.
I would like to understand the recommended Ignition Gateway/server sizing for this scale.
Could you please share your experience or recommendations regarding:
Recommended CPU/core count
Recommended RAM
Storage requirements
Practical tag capacity per Gateway
Impact of different OPC UA scan rates
Impact of alarming and tag historian
Perspective client load
Whether a single Ignition Gateway is suitable for ~100,000 tags or whether a distributed architecture is recommended
Any benchmark or reference architecture available for similar-scale systems
The intention is to understand the infrastructure required before proceeding with performance validation.
I've got a client running a standard redundancy setup with around 150k tags between the 2 main PLCs with another 60 or so other devices reading about a dozen tags from each. The other devices are modbus and ABB TotalFlow devices. The main PLCs were ControlLogix L83 processors in the past but new plants are now the L908TS models. We're using our own AOIs and UDTs for these projects and have optimized the comms to reduce load on the processors. They're running 8 core VMs with 16GB of RAM on Windows Server 2025. I believe the host is using Xeon Gold processors but it's have to check. I've configured Ignition to use a maximum memory of 8GB and CPU usage sits around 5% with some spikes up to 10% here and there. We have anywhere from 4-6 Vision clients running at a time with occasional increases here and there depending on anyone remoting into the system.
We have a separate server running Postgres with TimescaleDB for the historian and other databases (audit log and alarm journal). We're historizing about 7,500 tags and store around 15M values per day. Performance of the historian has been great also.
Thanks for the details Michael. Just to clarify your architecture, with the standard redundancy setup, are you running two Ignition Gateway VMs (Primary and Backup), each with 8 cores and 16 GB RAM, plus a separate PostgreSQL/TimescaleDB server? Also, are the 4–6 Vision clients running on separate client machines/VMs, or are they only concurrent Vision client sessions?
Primary and backup are on separate hosts. Since Ignition doesn't do load balancing between redundant gateways, all clients are running off of the active server (typically the master).
2 clients are 2 redundant servers running ThinManager with thin clients running load balanced (1 per server but fail over to the other server if one RDS server goes down). These 2 clients are running 4K TVs with a quad window setup so they can have 4 windows pulled up at once (still a single client though). The other 2 clients are panel PCs in the plant.
The other variable clients beyond the 4 are myself when doing development and testing and other users on a separate RDS server where we allow any number of clients to connect. Usually during a startup we have 3 or 4 engineers or more monitoring the system from their laptops on this remote connection on a single client RDS server.
Once normal operations is in place the remote connection won't be used as heavily but still gets used daily.
One point to consider is what sort of scan rate are you expecting on average? A remote monitoring system with 30-60 sec poll times vs a local server acting as a plant HMI and 1 second poll times is a significant difference in compute power.
But generally speaking Ignition is relatively lightweight. We have 40k OPC tags so far and it's a rounding error in terms CPU usage.