Suggestions on Long-Term Stability Test Parameters

When Spark is running, the number of executors and the memory size have a significant impact on performance. Higher concurrency means better performance benefits. However, to ensure stable service running, you are advised to check the system resource usage and the memory usage of resident processes to determine the final parameter settings. You are advised to use 1/3 of the performance parameter value. For details about how to optimize performance parameters, see Kunpeng BoostKit for Big Data Machine Learning Algorithm Library Tuning Guide.