We're setting up a non-RAC Oracle 188.8.131.52 instance on an x86-64 Redhat Linux 5.7 platform. The ASM diskgroups will live on NetApp storage (currently set up as block storage but may switch to NFS in the future). The NetApp storage is new to us. Oracle recommends creating ASM diskgroups with a 4MB Allocation Unit Size for enhanced performance. The "AU_SIZE" default is 1MB and this is what we've used in the past with 184.108.40.206 Oracle ASM diskgroups deployed on an HP EVA SAN.
Oracle caveats the 4MB recommendation by saying, to take advantage of the 4MB au_size, the operating system (OS) I/O size should be set to the largest possible size:
To anyone here who has deployed ASM diskgroups on NetApp, is this something we set in Redhat, somewhere in the Netapp device's OS, or both? Would love to use a 4MB AU_SIZE if it's a huge performance increase. But don't want to create performance problems if there would then be a mismatch with the physical storage parameters. Does NetApp have an AU_SIZE recommendation? I've not seen this item mentioned in any NetApp for Oracle 11g Best Practice documents, the SnapManager docs, etc. If anyone could point to a specific NetApp doc with Allocation Unit Size recommendations that would be great.
Anyone? I've cross-posted to the official Oracle forums with the same question. One gentleman believes in sticking with the default Allocation Unit Size of 1MB unless there is something special about the system (due to potential bugs, etc). An analogy was made with deviating from the standard 8K (8192 byte) Oracle Database block size. It can be done but is rare:
Would still love to hear from people using NetApp devices for Oracle ASM Diskgroups. What does NetApp recommend? And if you've deviated from the 1MB default AU_SIZE, e.g. say to the 4MB Oracle Corp recommendation, did you notice any appreciable performance gain?