That certainly sounds unusual. 

Are you seeing this when the JSON source files are identical? 

For the 1.14 build, the operator overview of that CTAS query's profile, you 
should be able to what operator is consuming the most time. 

There isn't any direct change in 1.14 that would explain the slowness that you 
seem to be experiencing. My hunch at the moment is that the memory settings are 
different between the 2 builds. 
There was a change introduced in 1.13 that changed the memory behavior 
(https://issues.apache.org/jira/browse/DRILL-5741). Most likely, your 10sec run 
is dragging into 5 minutes because of limited heap memory.

~ Kunal
On 8/13/2018 11:15:26 AM, Mehran Dashti [ BRCO - PD ] 
<[email protected]> wrote:
Congratulations for drill 1.14
But I see a very annoying slowness in ctas json to parquet in recent releases.
Specifically in 1.14 I see that it  seems to be a blocker bug.
5 minute takes to insert a json file to parquet, where in 1.13 it takes 10 
seconds.
 
 
 
Is it sth that I should do to get better results?
 
 
 
 
Best Regards,
 
      [LOGO1]
Mehran Dashti
Product Leader
09125902452
 

Reply via email to