What better way to configure memory for an Oracle RAC 11g R2?

Question

I'm setting up a new environment in Oracle RAC 11g but I'm wondering how best to use the memory feature.

Today I have two servers with 128GB of RAM 1 core CPU with 16 cores.

I deployed 64GB of SSD to Swap, as much as possible to be able to install and configure the S.O.

The environment is configured in two RHEL 6.9.

Today memory management is in automatic mode using the 64GB Swap.

Any tips on how to improve memory usage or pga and sga?

---

ASMM current configuration.

NAME TYPE VALUE ------------------------------------ ----------- ------------------------------ hi_shared_memory_address integer 0 memory_max_target big integer 0 memory_target big integer 0 shared_memory_address integer 0 show parameter sga NAME TYPE VALUE ------------------------------------ ----------- ------------------------------ lock_sga boolean FALSE pre_page_sga boolean FALSE sga_max_size big integer 80G sga_target big integer 80G show parameter pga NAME TYPE VALUE ------------------------------------ ----------- ------------------------------ pga_aggregate_target big integer 30G 

Answer 1

ASMM (automatic shared memory management) usually works best in oracle (imho). THat involves setting the sga and pga. Unless you are doing very large query/sorts your pga generally does not need to be nearly as big as your sga. Oracle can safely take up to 80% of your memory on a linux box, so you probably dont want to assign more than about 102G to oracle. Your swap does not need to be more than 16G for 11g and above (see oracle doc). Without knowing a lot about your workload, I would start with sga_target=90g, sga_max_size=90g, pga_aggregate_target=10G. You dont say if you are 11gR1 or 11gR2, but in either case use hugepages to improve CPU usage. If using 11gR2 set use_large_pages=ONLY to ensure hugepages are used. You can look here for further details on hugepage settings: https://dbakerber.wordpress.com/2012/03/14/configuring-hugepages-for-oracle-on-linux/ 90G of hugepages would mean set vm.nr_hugepages to 46085, which gives a small buffer for the setting. I would use the SSD for redo logs.