Why Raw disks in OPS ?
What is the reason for having Raw disks for parallel server ?
Is there any specific reason behind this ?
block contention, excessive pinging yadda yadd yadda.
OPS can almost be considered beta for rac. The overhead associated with OPS is so high that Oracle insists that you use RAW devices to help reduce that overhead.
In 9i we allow clustered file systems and the RAC code has been rewritten to be appreciably more effiecent than in OPS through the
marvelous cache fusion.
I'm stmontgo and I approve of this message
Well, AFAIK it's not because of "block contention, excessive pinging yadda yadd yadda and other owerhead". It's simply because the underlying OS don't allow simultaneous access to same filesystem devices from differen nodes. It's not because of Oracle, it's because of the OS. And it's all because of proper locking mechanism - at OS level, not at database level.
In 9i, clustered file system was introduced to overcome this, but it is still entirely implemented at OS level. Database is only something that's sitting on top of that.
And BTW, there is at least one operating system (at least the one that I'm aware of) that doesn't require RAW devices fo Oracle Parallel Server, for any release of Oracle OPS form Oracle7 upward! So this proves that OPS doesn't require RAW devices by its design - it can function as efficiently on ordinar file system if only the underlying OS allow it.
ASCII a stupid question, get a stupid ANSI
24 hours in a day .... 24 beer in a case .... coincidence?
Click Here to Expand Forum to Full Width