Why do some statement types take longer than others?
Example: ALTER takes significantly more time to execute than DROP
Depending on the type of call, the distributed nature of the call and the resource cost of the different types of executions are different.
In the above example, ALTER requires significantly more resources to execute as it scans and edits metadata in the Okera Catalog. While there are no changes to the underlying data made by the Okera Data Access Service (ODAS), the time needed to make this change will vary based on the number of columns in a table and what kind of alternations you are requesting.
DROP does not require any metadata modification, but still scans the selected table metadata before deleting it from the Okera Catalog. As a result, execution time is relatively low compared to ALTER, but is still dependent on size of table.
ODAS captures all timestamp information in the ODAS Audit logs.
Note: Timestamps in the Audit logs reflect the time the request was received by the planner (request_time) and when it was completed/failed by the planner (start_unix_time). The time in between reflects actual execution time.
Additional information can be found here on our Auditing documentation.