Protocol Benchmarking - OSI-Pi
The numbers in the tables below represent the ideal throughput for the data pump use case (data traveling from a protocol adapter to the Predix Edge MQTT Broker to the Cloud Gateway to Time Series).
The tests were executed in a controlled environment with one adapter running at a time, under ideal network conditions with a local time series simulator. The rate was deemed successful if there was less than a 1 percent loss rate over the course of a multi-hour test. Based on the testing, data loss rates dramatically increase as tag counts pass these thresholds.
| VM | Processors | RAM (GB) | Disk Space (GB) |
|---|---|---|---|
| VM (small) | 1 | 1 | 5 |
| VM (medium) | 2 | 4 | 20 |
| VM (large) | 4 | 8 | 20 |
| Environment | Period (seconds) | Acceptable Tag Rate |
|---|---|---|
| VM (small) | 1 | 150 |
| 10 | 2000* | |
| VM (medium) | 1 | TBD |
| 10 | TBD | |
| VM (large) | 1 | TBD |
| 10 | TBD | |
| Predix Edge Gateway 3002 | 1 | 525 |
| 10 | 2000* |
* This value represents the limit to which testing was conducted, rather than a limit that indicates increased tag counts would result in a 1 percent loss rate.
The number of tags per adapter does not scale with the device’s performance characteristics. It is recommended to add new adapters to support higher tag counts along with corresponding physical hardware to support the additional applications.