Benchmarking a System Built from the Static Structural Test Reference Architecture

To create performance benchmarks for the Static Structural Test Reference Architecture design pattern, NI built a physical test system according to the pattern and tested the system for viability and reliability, using the following tests.

  • Long-Term Test Results—Assessed the performance of the benchmark system when continuously logging data without producing errors.
  • Channel Count Parametrized Test Results—Found the highest sample rates the benchmark system could achieve at different channel counts.
  • Shunt Calibration and Null Offset Test Results—Assessed the time to complete a shunt calibration and null offset.

    • Use the results of the benchmark system tests as reference when building your own system from the Static Structural Test Reference Architecture.

    The Static Structural Test Benchmark System

    The following figure depicts the composition of the Static Structural Test benchmark system built by NI.

    Figure 13. Diagram of Benchmark System Built by NI

    1378

    1. cDAQ-9189: 8-Slot, TSN-Enabled Ethernet CompactDAQ Chassis
    2. cRIO-9805: 4-Port, 802.1AS Ethernet Switch Expansion Module for CompactRIO
    3. Cisco RV340: Router
    4. Lenovo ThinkStation P330 Tower Gen 2: Server Computer
    Note To stress scalability and performance, NI configured the benchmark system with over 2,000 channels—the top-end channel count suggested in the Static Structural Test Reference Architecture.

    The 32 cDAQ-9189 8-Slot chassis in the benchmark system are configured with a total of 256 modules and a channel composition that represents a common static structural test application.

    Table 18. Channel Composition and Sample Rates for the Benchmark System Built by NI
    Number and Type of Channels Number of Modules Module Models Sample Rate
    1,944 quarter-bridge strain gauges 243 NI-9235 and NI-9236 100 S/s
    32 voltage 8 NI-9215 100 S/s
    64 temperature 4 NI-9213 1 S/s

    Conditions

    The benchmark test results were found under the following conditions unless otherwise noted.

    • Static Test Software Suite 1.2 installed
    • Server computer: Lenovo ThinkStation P330 Tower Gen2
      • Intel Xeon E-2286G CPU 4.0 GHz, 6-Core
      • 64 GB RAM
      • 1 TB SSD
      • Gigabit Ethernet adapter (x2) (one dedicated for instrumentation network)
      • Windows 10 64-bit
    • External RAID drive: Oyen Digital 4TB SSD MiniPro RAID v3 (2 TB, RAID 1 configuration)
    • External DHCP-enabled router: Cisco RV340 router
    • Google Chrome browser
    • cDAQ-9189 chassis with cDAQ firmware version 20.0 installed (available through NI-DAQmx cDAQ Firmware 21.0)
    • System components connected according to the Static Structural Test Reference Architecture design pattern, as shown in the previous figure

    Long-Term Test Results

    The Long-Term Test assessed the performance of the benchmark system when continuously logging data without producing errors.

    Conditions:

    • 2-week test time
    • 2,048 channel count
    • 100 S/s sample rate for strain and voltage channels
    • 1 S/s sample rate for thermocouple channels
    Performance

    CPU usage

    20% to 70%

    Memory usage

    15% to 30%, 2,000 MB to 6,000 MB

    Network utility

    17% to 18%

    Storage utilization

    Total utilization

    2.15 TB

    Utilization by file type

    337 technical data management streaming (TDMS) files at ~5.91 GB each

    337 TDMS index files at ~620 MB each

    Channel Count Parametrized Test Results

    The Channel Count Parametrized Test found the highest sample rates the benchmark system could achieve at different channel counts. The highest sample rates, reported in the following section, were achieved for 30 minutes without producing errors and within the following parameters:

    • < 80% memory usage
    • < 95% CPU usage
    • < 80% network utility
    Highest sample rates by channel count
    2,048 channels

    Maximum sample rate

    3 kS/s

    CPU usage

    36% to 50%

    Memory usage

    28% to 30%

    Network utility

    ~22%

    1,160 channels

    Maximum sample rate

    10 kS/s (NI-9235 and NI-9236 have a maximum sample rate of 10 kS/s)

    CPU usage

    40% to 60%

    Memory usage

    34% to 40%

    Network utility

    ~38%

    448 channels

    Maximum sample rate

    10 kS/s (NI-9235 and NI-9236 have a maximum sample rate of 10 kS/s)

    CPU usage

    20% to 50%

    Memory usage

    20% to 30%

    Network utility

    ~15%

    Shunt Calibration and Null Offset Test Results

    Time to complete shunt calibration on 1,944 channels

    5 minutes

    Time to complete null offset on 1,944 channels

    5 minutes