Best Practices
- Last UpdatedMay 20, 2026
- 2 minute read
- PI System
- PI System Connector 3 3.0
- Connectors
Do not put the connector and broker on the same machine. The broker and the connector should be on dedicated machines with no other applications (such as a SQL Server, PI Server, AF Server, etc.).
Verify that the AVEVA™ PI System Connector 3 (PSC3) and PI System Connector Broker are not competing with Source and Destination clients. If you have an application (PI Vision, Datalink, or something custom) that makes expensive queries to your source or destination PI Server, this can impact performance.
If you have a PI Server collective, set the affinity for AVEVA™ PI System Connector and PI System Connector Broker to a node that is less overloaded/used.
System Sizing Best Practices
|
Category |
Guideline |
|---|---|
|
General |
The following recommendations given are based on a typical deployment that handles 20,000 events per second and 500,000 data points. It provides sufficient assurance of acceptable performance for smaller systems at peak load times (AF exports, History Recovery, bulk edits, etc.) |
|
Virtual Processor minimum requirements |
Minimum of eight virtual processors |
|
RAM minimum requirements |
Must be at least 1.3x size of the largest AF database the connector is expected to process. Although you may not require this much for normal operation, if you have less RAM, the system will struggle at peak load times. The following are minimal requirements recommended for optimal performance:
|
|
Event Frames |
Additional 4GB RAM for every 4 million Event Frames. |
|
Large AF Database |
Additional 2 GB Ram for every 100,000 elements. |
|
Large PI Point Count and Snapshot Rate |
For systems that are expected to exceed deployment parameters stated above, we recommend using a minimum of 16 virtual processors and at least 64 GB of RAM. |
Make Health AF Element, pointing to customer created health tags.
You will need a watchdog tag on the Source PI Data Archive– one that updates very frequently – to monitor updates are being replicated. We recommend a tag is added to the source PI Data Archive that updates once per second (or customer preference).
Create an element with a reference to the newly created tag bound to an attribute so the custom element is replicated and can be monitored on the destination.