> Rows that are being returned are most likely not yet purged. Rows that have been deleted won’t be returned in the result set. SELECT statements default to being executed on the latest epoch, which won’t include the deleted records. Whether or not deleted records have been purged is irrelevant, sin...

Systems tables are really views on top of Data Collector tables. So the time span for the data in the system table is dependent upon the retention of the data in the DC tables. You can see the current effective retention for each table on each node in the DATA_COLLECTOR table. You can increase the r...

Already answered in the HP Vertica Community.

See this blog entry on the new SC1 volumes.

http://eng.localytics.com/amazon-ebs-sc ... pressions/

Hi Janica,

There is probably some other operator that has a long "clock time (us)" counter_value in the execution_engine_profiles data. You can correlate the profiling data to the explain plan using the path ids.

--Sharon

One clarification on the previous reply. That example isn’t using a flex table - it’s a CREATE TABLE rather than CREATE FLEX TABLE. It’s creating a vmap for the json, and using the MAP functions to access it. Creating a flex table gives you more transparent mechanisms for accessing the json keys as ...

So then AT EPOCH LATEST is already effectively in use with READ COMMITTED - no need to use that. It would be worth understanding the actual source of the slowdown when you have the CDC running. Using AT EPOCH LATEST isn’t the solution.

—Sharon