Performance: OPC UA Data Modeling

This project collects 100,000+ OPC UA tags, models the data as 10,000+ pumps, and sends the data to MQTT once a second.

KEPServerEX is used as the OPC UA server and is installed on a separate networked machine. The Intelligence Hub is installed on a single machine with the MQTT Broker enabled.

See below links to download the project files to run this test in your own environment.

Project Details

• Number of Pumps: 10,240
• Number of Tags per Pump: 9
• Number of Tags: 92,160
• Flow publish rate: 1 second
• Number of MQTT Connections: 3
• Number of Flows: 5

System Specs

• AWS EC2 t2.2xlarge
  • vCPU: 8
  • Memory: 32GB

Performance Results

• Write payloads/second: 10,296 vs 10,240
• Write attributes/second: 123,560 vs 122,880
• CPU: 46.3%
• Memory: 84%
• Aggregated results: aggregate_results.json
• Raw results: raw_results.json

Testing Notes

• Results averaged over 10 minute time window
• Attributes/second includes internal _name, _model, _timestamp fields
• Write results are greater than expected because of buffering/bursts. This should be interpreted as meeting expected results.
• 3 MQTT connections are used to spread the write load across 3 threads

Steps to Reproduce

• Install the latest version of the Intelligence Hub
• Install KEPServerEX on the same or a different machine to match this test
• Download and load the kepserverex_pumps.opf KEPServerEX project and ensure OPC UA is enabled
• Download and load the intelligencehub_pumps.json Intelligence Hub Project
  • Change the OPC UA Connection settings to connect to your OPC UA Server
  • Enable the local MQTT Broker under Settings
• Confirm connectivity to the OPC UA Server and MQTT Broker is working
• Enable the Flows