After upgrading to hm-2022.2, our performance analysis dashboards take 45-60 seconds to load, sometimes timing out completely. The OEE dashboard and production efficiency charts are particularly slow. We have about 6 months of historical data with 12 production lines reporting metrics every 5 minutes. Before the cloud migration, these dashboards loaded in under 5 seconds on our on-premise deployment. The data is there and accurate when it finally loads, but the wait time is frustrating our production managers who need quick access to real-time metrics during shift meetings. Is this normal for cloud deployments or is there a configuration issue?
Dashboard performance in cloud deployments depends heavily on how the data aggregation is configured. Check if your dashboards are querying raw data versus pre-aggregated tables. In Administration → Performance Analysis → Dashboard Settings, look for ‘Data Aggregation Level’. If it’s set to ‘Real-time Raw Data’, that’s your problem - it’s querying millions of 5-minute interval records. Change it to ‘Hourly Aggregates’ for historical views.
Real-time dashboard queries hit the performance_metrics table which can grow massive with 12 production lines at 5-minute intervals. That’s 3,456 records per day per line, or about 41,000 records daily total. After 6 months you’re at 7.4 million records. Check if the table has proper indexes on timestamp and production_line_id columns. Also verify the API gateway timeout settings - default 30 seconds might be too short for complex aggregation queries. We increased ours to 60 seconds for performance dashboards specifically.
Mike, how do I check the database indexes? I’m not seeing index information in the admin interface. And regarding the API gateway timeout that cloud_performance_eng mentioned, where is that configured? I found the caching settings and dashboard cache was disabled - I’ve enabled it now with the recommended TTL values.
Sarah, I checked and the Data Aggregation Level is already set to ‘Hourly Aggregates’ for dashboards showing data older than 24 hours. The slow performance affects even current shift dashboards though, which need real-time data. Is there a way to optimize the real-time queries?