By JeffG-Oracle on Jul 17, 2013
Due to the large amount of data within our system we are plannig to implement RAC to improve performance. We have already recommended other performance guidelines from Oracle like reordering columns, partitioning sales_data table etc. For further performance improvement since we expect more stores and items being introduced which would increase data we are planning to go for RAC as it is mentioned in the Demantra Installation Guide that RAC is supported in Demanra.
We would like the below information before proceeding with this major step:-
1.) For Distributed Engine Scaling - The Oracle Doc ID 468688.1 states that "The number of engines/machines defined in the Engine Administrator should not exceed 3 times the number of database CPUs". So we want to know in case of RAC will the above calculation be based on cumulative CPU nodes of all database RAC nodes? I.e when using RAC can we set the total number of Engines upto 3 times the cumulative CPU nodes of all database RAC nodes.
Answer 1: For Distributed Engine Scaling -- using the cumulative CPU nodes is not quite right for RAC, since there will be extra interconnect overhead in a RAC system. So you can try using the "3 times CPU's" but you might have to scale it back based on empirical performance.
2.) Using a Distributed Engine with one master and multiple slave machines. Is it possible to point different slave engines to different RAC nodes?
Answer 2: Is it possible to point different slave engines to different RAC nodes? No. The RAC will automatically load balance when you connect using the SCAN addresses.