| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Mailing Lists -> Oracle-L -> Re: Oracle manual : Greater O/S block size......
Aldi,
If you are running a "vxfs filesystem", the filesystem blocksize overrides the default O/S block size of 512 bytes, from the perspective of I/O operations to that "vxfs filesystem". It is probably 8k if you took the default values while configuring your vxfs filesystem.
This means that your db_block_size should be atleast 8k, given the OLTP/Batch nature of your application. If you are using "Quick I/O", then you may even want to set db_block_size = vxfs filesystem blocksize to minimize overhead during mapping. If you are running a lot of batch jobs through the day, then you may even go to a 16K db_block_size. Just make sure that "block-level contention" parameters like FREELISTS, INITRANS etc. are configured appropriately.
The term "operating system I/O size" refers to the "chunksize" that the O/S uses to perform an I/O operation. On most flavors of Unix it is 64K. I believe, in more recent versions of Solaris, this is configurable and can be set to 128K or 256K. This however should be set keeping in mind the limitations of your hardware. After configuring that, the degree of striping (stripewidth) that you deploy in your logical volumes and the values you set for init.ora parameters like db_file_multiblock_read_count and hash_multiblock_io_count should be aligned with the "O/S I/O size" or "chunksize".
It is my understanding that you cannot change the "O/S blocksize" after you have created your filesystems, as it refers to the "blocking factor" within a filesystem. But if that can be changed, someone please let us know.
Best Regards,
Gaja.
"Opinions and views expressed are my own and not of Quest"
 |
 |