LVM and VxVM

Re: VXFS/JFS block size/frag size vs database block size

 
Fredrik Uddin-Backman2
Frequent Advisor

VXFS/JFS block size/frag size vs database block size

Hi

Whats the take nowadays regarding block size in HPUX. I hear opinions about newer HPUX dont care "as much" about block size and will group I/O together in large chunks regardless of choosen block size. If running database (Progress v10) with 8k pagesize you should in my book choose 8k for your fs blocksize also but I see that newer HPUX have 1k as default if volume<8TB? What is the drawback if choosing the larger size even for smaller volume?
2 REPLIES 2
Emil Velez
Honored Contributor

Re: VXFS/JFS block size/frag size vs database block size

Default for large vxfs filesystems is 8k..

Remember that most of your I/Os are done to/from memory using buffer cache unless you use the mincache paramter or are running 11.31 with large files.

Alot depends on what you have buffer cache set for and what type of storage you have.

Fredrik Uddin-Backman2
Frequent Advisor

Re: VXFS/JFS block size/frag size vs database block size

Storage is one or two XP24000 depending on mirror requirements.

Regarding buffer cache, I have dbc_max_pct=4% and dbc_min_pct=4%