Hi,
I have 8 mysqld with 2 datanodes (multithreaded with 8 max execution threads), Diskless=1, RedoBuffer=128M. Our application is generating 5000 updates/inserts per second using JDBC batching but we can't seem to go over 1500 updates/inserts per second.
The cluster is all running on the same 16-way server box with 96GB of RAM including the load generating application.
What are your recommendations to get this running quicker?
Should I be adding more data nodes or instead switch to a lower level API like ClusterJ to gain the desired write performance?
I have 8 mysqld with 2 datanodes (multithreaded with 8 max execution threads), Diskless=1, RedoBuffer=128M. Our application is generating 5000 updates/inserts per second using JDBC batching but we can't seem to go over 1500 updates/inserts per second.
The cluster is all running on the same 16-way server box with 96GB of RAM including the load generating application.
What are your recommendations to get this running quicker?
Should I be adding more data nodes or instead switch to a lower level API like ClusterJ to gain the desired write performance?
