Decision Insight 20200928 Save PDF Selected topic Selected topic and subtopics All content How to capitalize the CPU usage? If your CPU cores are under-used, take better advantage of your hardware by tuning the following parameters. Increasing the following parameters will also increase the consumed memory. Do not increase these if your memory usage is high. Please follow the steps below to tune your CPU usage 1- Identify your performance concern The default parameters are suitable for most recommended configurations. Change them only if you want to alleviate a performance concern. Here are the top CPU consumer components: Component Effects Dashboards All dashboards are slow to display data (if a single dashboard is slow, review its configuration). Dashboards become slower as more users connect. Live computing Dashboards sometimes show <Not set> values for computed attributes. Monitoring screens show that the average duration of an attribute is close or above computing rhythm. Computing correction Late data is not taken into account in a timely manner. See also Computing options. Monitoring screens show a very high number of corrections and this number barely decreases. 2- Check your component health Component Effects Dashboards Do not add too many dimensions by pagelet: 6 dimensions should be the maximum with no more than one transaction dimension. For a Search dashboard: Prefer several dashboards by domain instead of a single one with many criteria. Use a large refresh rhythm; this type of dashboard typically does not require refreshing data every minute. Check the Load data on demand option in dashboard settings. Consider using a Clustered deployment if you have many users connecting at the same time. Live computing Use a computing rhythm as large as possible to avoid useless pressure. Delete useless attributes. For more information, see How to find unused attributes in an application. Computing correction Minimize corrections by preventing late absorptions in data integration by configuring the Active recomputing period option as described in Computing options. Purge data to get rid of obsolete data. This will alleviate data query. 3- Tune parameters Parameter Description Default com.systar.krypton.scheduler.poolSize Maximum number of Live computing tasks executed in parallel. 10 com.systar.krypton.scheduler.correctionPoolSize Maximum number of computing correction tasks executed in parallel. 5 com.systar.tungsten.maxConcurrentPlanExecution Maximum number of queries executed in parallel.The queries can come from computing or dashboards. 20 Keep in mind that com.systar.tungsten.maxConcurrentPlanExecution must be greater than com.systar.krypton.scheduler.poolSize + com.systar.krypton.scheduler.correctionPoolSize to give space to others queries (dashboards, data integration queries, ...) Component to improve Operations Dashboards Increase only com.systar.tungsten.maxConcurrentPlanExecution. Live computing Increase com.systar.krypton.scheduler.poolSize and com.systar.tungsten.maxConcurrentPlanExecution accordingly. Computing correction Increase com.systar.krypton.scheduler.correctionPoolSize and com.systar.tungsten.maxConcurrentPlanExecution accordingly. Related Links
How to capitalize the CPU usage? If your CPU cores are under-used, take better advantage of your hardware by tuning the following parameters. Increasing the following parameters will also increase the consumed memory. Do not increase these if your memory usage is high. Please follow the steps below to tune your CPU usage 1- Identify your performance concern The default parameters are suitable for most recommended configurations. Change them only if you want to alleviate a performance concern. Here are the top CPU consumer components: Component Effects Dashboards All dashboards are slow to display data (if a single dashboard is slow, review its configuration). Dashboards become slower as more users connect. Live computing Dashboards sometimes show <Not set> values for computed attributes. Monitoring screens show that the average duration of an attribute is close or above computing rhythm. Computing correction Late data is not taken into account in a timely manner. See also Computing options. Monitoring screens show a very high number of corrections and this number barely decreases. 2- Check your component health Component Effects Dashboards Do not add too many dimensions by pagelet: 6 dimensions should be the maximum with no more than one transaction dimension. For a Search dashboard: Prefer several dashboards by domain instead of a single one with many criteria. Use a large refresh rhythm; this type of dashboard typically does not require refreshing data every minute. Check the Load data on demand option in dashboard settings. Consider using a Clustered deployment if you have many users connecting at the same time. Live computing Use a computing rhythm as large as possible to avoid useless pressure. Delete useless attributes. For more information, see How to find unused attributes in an application. Computing correction Minimize corrections by preventing late absorptions in data integration by configuring the Active recomputing period option as described in Computing options. Purge data to get rid of obsolete data. This will alleviate data query. 3- Tune parameters Parameter Description Default com.systar.krypton.scheduler.poolSize Maximum number of Live computing tasks executed in parallel. 10 com.systar.krypton.scheduler.correctionPoolSize Maximum number of computing correction tasks executed in parallel. 5 com.systar.tungsten.maxConcurrentPlanExecution Maximum number of queries executed in parallel.The queries can come from computing or dashboards. 20 Keep in mind that com.systar.tungsten.maxConcurrentPlanExecution must be greater than com.systar.krypton.scheduler.poolSize + com.systar.krypton.scheduler.correctionPoolSize to give space to others queries (dashboards, data integration queries, ...) Component to improve Operations Dashboards Increase only com.systar.tungsten.maxConcurrentPlanExecution. Live computing Increase com.systar.krypton.scheduler.poolSize and com.systar.tungsten.maxConcurrentPlanExecution accordingly. Computing correction Increase com.systar.krypton.scheduler.correctionPoolSize and com.systar.tungsten.maxConcurrentPlanExecution accordingly.