It's the memory latency, stupid
CPUs can appear to be at 100% even if they're not fully utilized due to memory latency, which is much slower than internal CPU operations. CPU caches are designed to mitigate this issue, but their small size is a limiting factor. Database performance engineers often focus on cache utilization to optimize performance. GROUP BY aggregations can help understand the impact of cache utilization.
Read the full article at pivotlake.io →