rbalamohan opened a new issue, #6726:
URL: https://github.com/apache/iceberg/issues/6726

   ### Apache Iceberg version
   
   1.0.0
   
   ### Query engine
   
   Hive
   
   ### Please describe the bug 🐞
   
   Context: 
   ======
   Split computation in iceberg takes 3-5 seconds (in certain queries more than 
that); This is significantly higher when compared to regular tables, which were 
in the order of 100s of milliseconds. BI kind of query latencies are impacted 
due to split computation runtimes.
   
   Issue:
   =====
   Dynamic partition filtering is a powerful optimization to reduce the number 
of partitions being scanned at runtime. In hive, it creates the filter at 
runtime and pushes it to downstream vertices at runtime. 
   
   In iceberg, split computation in IcebergInputFormat happens via 
scan.planTasks. Filters if any are applied later. This causes metadata for all 
partitions to be read and then applies the filter causing computatioin delays. 
It will be good to figure out if the filtering can be applied earlier or in 
optimal manner.
   
   
https://github.com/apache/iceberg/blob/dbb8a404f6632a55acb36e949f0e7b84b643cede/mr/src/main/java/org/apache/iceberg/mr/mapreduce/IcebergInputFormat.java#L147
 
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to