RE: A VLDB contest
- From: Tanel Poder <tanel.poder.003@xxxxxxx>
- To: kevinc@xxxxxxxxxxxxx, "'Bobak, Mark'" <Mark.Bobak@xxxxxxxxxxxxxxx>, "'ORACLE-L'" <oracle-l@xxxxxxxxxxxxx>
- Date: Wed, 30 Aug 2006 21:43:36 +0800
Also, when your server process needs metadata for an ASM allocation unit
which is not already cached in your databases memory, it needs to do an OCI
call to ASM instance. This in turn will make an IPC call to ASM instance,
possibly spawning a new ASM process.
And then ASM instance will read the metadata from disk, and pass it pack to
the waiting server process, then the server process has to get onto CPU
again to continue processing (for example loading rows into a table) - once
it requires a next ASM allocation unit for itself...
Also, no FUD intended here. The few small to medium databases what I've (had
to) set up with ASM have performed ok..
Tanel.
_____
From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx]
On Behalf Of Kevin Closson
Sent: Wednesday, August 30, 2006 07:39
To: Bobak, Mark; ORACLE-L
Subject: RE: A VLDB contest
Right, the ASM instance gives the pointer array info to the
consumer instances where it is cached in the shared pool.
I've not seen one iota of discussion on this list about how
Oracle maps a data block to an extent within an ASM diskgroup.
Must not be an interesting topic?
It is indeed using these pointers. Such mapping is never free. Volume
managers have always had to do that--but have always stored
such mappings in kernel memory. No FUD intended here. I just
wanted to get people curious about the topic.
Other related posts:
