Druid select query performance

What areas would you recommend addressing when trying to speed up Druid query response times? We have a select query that is returning 500000 records and it is taking up to 35 seconds for the Druid broker to complete.

Is this something query cache could speed up? Is it a matter of tuning the configuration better? Is it also likely to be a problem with the query or a problem with segments size and granularity? Thank you

Hi Chris,
If you have large number of small segments, you might want to do compaction first so that you will relatively smaller number of optimized segments (300MB to 900MB each) which will help in query performance.

By any chance, did you try scan instead of select to see whether it helps or not?



thanks so much for the response siva.
apparently we can’t do a scan.

i will look into the compaction issue.


Hey Chris,

If the reason you couldn’t use the Scan query was its (previous) lack of time ordering support, check out Druid 0.15.0. It added time ordering support for the Scan query, which is much more efficient than the Select query.

I will find out if this is the case. Thank you very much for your support as always.