Different block sizes.

[ron]

I've just started working on a moderately large (350 Gb) DB, OE 10.2B on Linux.

I notice that the DB block size is 4K, the BI block size is 16K and the AI block size is 8K. My query relates to the different sizes for AI and BI. I know I read somewhere that it's recommended to have them the same. Is there any fallout from them being different?

Ron.

 

[RealHeavyDude]

Nope. Usually the block sizes for the BI and AI are deliberately at their max for performance reasons. The BI and AI allow for bigger block sizes then the database extents. As a rule of thumb they should always be equal or a multiple of the files system block size their are residing on. If they are smaller than the file system block size then you will get a serious performance penalty. In an ideal world you would have distinct file systems for the BI, the AI and the database that have their block size set matching the block sizes you defined accordingly.

Heavy Regards, RealHeavyDude.

 

[Rob Fitzpatrick]

FWIW, I have been told to keep BI and AI block sizes the same, although I don't have RHD's level of experience.

Check that you don't have your BI cluster size at the default as that will constrain performance.

If you have a non-production database with reasonable application activity (test, QA, etc. ) and the time to do some benchmark testing, you may see benefit in moving to an 8K DB block size in terms of buffering performance and general application performance (i.e. less physical I/O for a given amount of logical I/O). That's assuming you're starting from a position where you're reasonably tuned, in terms of your structure: Type II areas, good RPB settings, etc.