RE: A VLDB contest

 >>>If you enable 10046 trace while doing the scan, you should 
>>>see your server process occasionally waiting on SQL*Net 
>>>message to/from client events against the ASM instance, 
>>>while it's scanning through the table. These are the ASM 
>>>metadata fetches.
>>>

right. I was looking at some stats from one of our buddies over
on the Oak Table and he showed the extant array values from all 4
instances on a 4 node cluster. Instance 4 had 10x more ASM extent
array size than the largest of the other 3 instances. I'm sure
we've all measured the failover impact of being
cache-cold on over 1GB of extent array cache that has to be
fetched over 1Gb right in the event of a RAC failover? What if that
forces a shared pool extention? What if that causes other shrinkage?
Hmmm...

I started testing this very aspect yesterday so the questions are
rhetorical :-)
 
--
http://www.freelists.org/webpage/oracle-l

Other related posts: