inverted index will not help

@Kishore G I tried benchmarking the same query with a range index on

start_time_millis

but it didn’t really help (the

timeUsersMs

were more or less the same). However,

numEntriesScannedInFilter

came down from 25000000 to 2249975.

timeUsedMs

is around 500ms. Can you do anything else to bring this number down?

Is there something special with start Millie and end Millie

It’s just the millis since the Epoch.

What about the query granularity

We are querying on the millis level as well (if this is what you meant). Example: Select api_id, service_name, service_id, api_name, COUNT(*) FROM rawServiceView WHERE tenant_id = ‘__default’ AND ( api_id IS NOT NULL AND start_time_millis >= 1623070770000 AND start_time_millis < 1626271772000) GROUP BY api_id, service_name, service_id, api_name ORDER BY PERCENTILETDIGEST99(duration_millis) desc limit 10000

Got it. You can make it faster by increasing the number of ranges in range index

Hey Kishore thanks for the quick help on this. I am not sure what you mean by increasing the number of ranges in the range index. We have just enabled it on the single column (start_time_millis) and are querying on that. Is it further configurable?