Pushing query execution to DataNode

[Kok Fong Lau]

I'm trying to run the following query in MySQL Cluster:

select * from (
select
a.*,
@row_num := if (@curr_score = score, @row_num + 1, 1) as rn,
@curr_score := score
from
score_result a
order by score, created_dt desc
) where rn = 1;

The score_result table has 5 million records.

Running the above query, we realized that the sql node is requesting the data from the data node then processed within the sql node.

This would explain why the query is slow.

For joins there's a ndb_join_pushdown flag which let's the data node perform the join filter first then only send to the sql node.

Is there an equivalent parameter for the above query such that I want the data to be processed in the data node then only extract the records where rn = 1;