It’s working okay with a 30 day rolling average… (every day we truncate older rows) we read from it to generate on demand status for per-second performance of tasks in a big job processing engine. But long term we want to have all the data available for historical analysis, trend analysis etc.