Details
- 
    Bug 
- 
    Status: Closed (View Workflow)
- 
    Major 
- 
    Resolution: Duplicate
- 
    None
- 
    None
- 
        2021-4
Description
It gets much worse if a UNION ALL is involved.
If you do 
select count(c) from (select * from t1) q;
select count(c) from (select * from t2) q;
and compare it to 
select count(c) from (select * from t1 union all select * from t2) q; the difference in elapsed and CPU is around 10x (second query is 10x worse than the sum of the first two).
Again, plans and I/O stats are identical (second is the same as sum of the first two).
Attachments
Issue Links
- duplicates
- 
                    MCOL-4901 Allow pushdown of queries involving UNIONs in outer select to ColumnStore -         
- Closed
 
-         
- is caused by
- 
                    MDEV-25080 Allow pushdown of queries involving UNIONs in outer select to foreign engines -         
- Closed
 
-         
- is duplicated by
- 
                    MCOL-4563 Investigate and correct a CPU / elapsed anomaly with UNION -         
- Closed
 
-         
- relates to
- 
                    MCOL-4584 Significant performance degradation when UNION ALL is used in an outer select, compared to in a subquery. -         
- Closed
 
-         
- split from
- 
                    MCOL-4543 Locate and Remove CPU waste in PM1 on nested queries with aggregates -         
- Closed
 
-