Benchmarking NI Linux RT Application Performance

If your application has demanding performance requirements, you should create performance benchmark tests for your application to ensure it meets your performance requirements. Performance requirements are highly application dependent, and you should determine what performance metrics are important. Typical real-time application performance metrics are discussed below.

If migrating an existing application from PharLap, you should create performance benchmarks of your existing application before starting the migration process. Once you have migrated your application, you can profile your application performance on NI LinuxRT. NI R&D has profiled performance of NI LinuxRT versus PharLap, and there are several key performance takeaways to highlight:

• NI LinuxRT determinism is in line with Pharlap
• NI LinuxRT Ethernet and disk throughput are significantly improved from Pharlap
• General application-level benchmarks show comparable performance.
• For an example of application benchmarks on cRIO with NI LinuxRT, see the NI CompactRIO Performance Controller: Performance and Throughput Benchmarks.

Check Loop Execution Rate

Loop execution rate refers to the amount of time that is required to execute one iteration of your application’s main loop; eg can my application read 10 analog inputs in 100 m? To measure loop execution time, see the Example Finder > Toolkits and Modules >  Real-Time > Benchmarking > Benchmark Project example include with the LabVIEW Real-Time Module. Loop execution rate is highly correlated to CPU and memory usage, and these measurements are addressed in the next sections.

Check CPU Usage

The CPU usage is an important performance metric to monitor when evaluating application performance.
To measure CPU usage on LinuxRT, you can use common Linux tools such as top or htop or NI tools.

Check Application Throughput, Jitter and Latency

For an example of jitter measurement, see the RT Analysis Workspace example include with the LabVIEW Real-Time Module. While the general application-level benchmarks show comparable performance, we recommend profiling your application for jitter and latency if your application is sensitive to these metrics.

Check Memory Usage

Memory management and reporting on Linux devices differs significantly from targets running operating systems such as Windows or PharLap.

Check Network and Disk Throughput

As noted in the overview, NI LinuxRT Ethernet and disk throughputs are significantly improved from Pharlap. See below for sample test results from NI performance benchmark tests.

Figure 1: Comparison of network throughput test running on LinuxRT versus PharLap


Figure 2: Comparison of disk write throughput test running on LinuxRT versus PharLap