We have a cluster installation of 4 data nodes and 1 api node.
Cluster has one database with 1 big disk-less table with 8 million inserts per day.
Our server's memory allow to store data for 4 days. We have a cron job, which moves old data (4 days ago) to innodb table every night (8 million deletes).
After delete performance of select queries on that table degrades drastically.
For example query execution time with newly created table with 32 million rows is 1-5 seconds. And same query execution time after delete 8 millions is 90-100 seconds.
How to implement effective sliding window behavior in ndbcluster engine?
Cluster has one database with 1 big disk-less table with 8 million inserts per day.
Our server's memory allow to store data for 4 days. We have a cron job, which moves old data (4 days ago) to innodb table every night (8 million deletes).
After delete performance of select queries on that table degrades drastically.
For example query execution time with newly created table with 32 million rows is 1-5 seconds. And same query execution time after delete 8 millions is 90-100 seconds.
How to implement effective sliding window behavior in ndbcluster engine?