Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> pmap - anon pages
Hi,
This is a datawarehouse system , and parallel query is mostly used.
I am trying to locate the oracle pga memory allocation inside the
process in
order to size my systems memory requirements. I have read the documents
in
metalink but i believe i could not find the answer of the following
questions:
When i monitor an oracle process by using pmap , i can easily see
shared memory
, stack , heap , libraries. By using this output i issue some sqls and
check
pmap output .
example queries:
SELECT /*+PARALLEL(T,32)*/ * FROM TOLGA2 t order by 1
SELECT /*+noPARALLEL(T)*/ * FROM TOLGA2 t order by 1
i was told that oracle parallel query bypasses the oracle buffer cache
and
uses process memory instead.
Based on these backgrounds ,
1. For the parallel query , i have tried various sort_Area_size
parameters with
manual workarea management in order to see its effect in pga memory
allocation
(and pmap output ) . What i see in pmap output is , lots of anon
(anonymous
pages) are allocated instead of heap. But according to the documents
pga is
allocated inside the heap. (Anon page is a page that is not associated
with a
particular vnode, or file/directory )Is the data stored inside this
anon pages?
2. for the nonparallel query , i assume that the data blocks now ara
cached
inside the oracle sga which is in shared memory but i can see the same
anon
page allocation pattern here. Why does the process allocates anon pages
like
parallel query case although data is cached inside the shared memory
(and it
is also mapped in process)
Kind Regards,
tolga
Received on Thu Jul 28 2005 - 06:01:05 CDT