7.4. Number of Analytical Queries Executed per Unit of Time #
After allocating the computational resources required for executing analytical queries, calculate the total workload from analytical queries executed in parallel. This workload tends to peak once during business hours or when performing resource-intensive ETL operations.
Certain ETL operations can be scheduled to be performed at a specific time of day, whereas others — such as ad-hoc analytical queries — cannot.
Avoid OLAP workload peaks by evenly distributing the execution of analytical queries throughout the day and leaving a 30% margin for unexpected operations.
pgpro_axe processes analytical queries linearly as long as it has sufficient I/O resources, RAM, and CPUs. Once these resources are exhausted, the performance drops.
Recommendations for avoiding resource exhaustion:
eliminate bottlenecks in the Postgres Pro AXE cluster
reduce OLAP workload peaks
optimize the execution of analytical queries
redistribute ETL operations and ad-hoc analytical queries to a dedicated Postgres Pro AXE server, for example, to an asynchronous standby replica