Element Size - RAID 10 SAN
Wondering about the optimal configuration for a Raid 10 array, specifically, element/block/stripe sizes. Does this relate to my DB block size the same way it would with Oracle? Given the smaller the element size, the more efficient the distribution of data written, and that most writing logic will write to the same stripe multiple times if more than one page will fit (again, much more efficient) where do I find the happy medium? Thanks for any and all advice, Tam.
> Wondering about the optimal configuration for a Raid 10 > array, specifically, element/block/stripe sizes. > > Does this relate to my DB block size the same way it would > with Oracle? > > Given the smaller the element size, the more efficient the > distribution of data written, and that most writing logic > will write to the same stripe multiple times if more than one > page will fit (again, much more efficient) where do I find > the happy medium? In the middle ? > > Thanks for any and all advice, > Not sure about optimal but 16K sounds nice Paul Watson Tel: +44 1414161772 Mob: +44 7818003457 Web: www.oninit.com GO FURTHER with DB2 GET THERE FASTER with Informix. Attend IDUG 2007 San Jose, North America May 6-10, 2007 Visit http://www.iiug.org/conf for more information.
Tam O'Shanter wrote: > Wondering about the optimal configuration for a Raid 10 array, specifically, > element/block/stripe sizes. > > Does this relate to my DB block size the same way it would with Oracle? > > Given the smaller the element size, the more efficient the distribution of > data written, and that most writing logic will write to the same stripe > multiple times if more than one page will fit (again, much more efficient) > where do I find the happy medium? Unlike setting up an array for a filesystem, you want to keep the stripe block size fairly small. IDS always writes either a single page or a 'big block' of eight pages. You definitely want to minimize the number of times that an array block is written to and increase the chances that cache gather operations will be able to combine writes as often as possible. For a 2K pagesize I like 16 or 32K blocks and I think that would work for 4K & 8K pagesizes just as well if not better. For larger pagesized I'd want to experiment with increasing that a bit, but I haven't tried it. Art S. Kagel