How to make RMAN to use large pool

[richyip]

I have allocated 32MB for large pool. Whenever RMAN backup runs, I notice higher PGA memory usage. I read in Oracle guide, large pool is used for RMAN. But when I issue 'select * from v$sgastat where pool = 'large pool'', I saw see the free memory is still 32MB. So how can I make use of large pool for RMAN backup operations? What I need to set?

[adewri]

http://metalink.oracle.com/metalink/...T&p_id=73354.1

http://metalink.oracle.com/metalink/..._id=156826.999

HTH

[jodie]

Amar,
I read the articles you posted, and thought I'd test it out. My results were not what I expected.

Large Pool = 80M
Block Size = 16K
db_file_direct_io_count = 128
Files = 9
Backup_tape_io_slaves = true
Backing up directly to tape.

From the first article, my memory usage should be:
memory(input) = 4 * #files * db_block_size * db_file_direct_io_count
= 4 * 9 * 16*1024 * 128
= 72 Meg

SQL> select * from v$sgastat where pool ='large pool';

POOL        NAME                            BYTES
----------- -------------------------- ----------
large pool  free memory                  80000000

SQL> /

While the backup is running, it only used 278K of the large pool.

POOL        NAME                            BYTES
----------- -------------------------- ----------
large pool  free memory                  79721424
large pool  KSFQ Buffers                   278576

What am I missing?

Thanks!
Jodie

[vinodkmg]

how many channels are using???
whatz ur filesperset ????

in general we need to allocated 4 buffers per file.
each buffer is db_block_size * multi_block_read_count

for eg:

db_block_size=8 and multi_block_read_count=8 then 64k per buffer

64k * 4 * 2(channels)

This config works fine for me.Rman uses large_pool_size properly

regards,
Vinod

[jodie]

Vinod,
My Filesperset is not set, so it defaults to all my files (9). I am only using 1 channel.

Also, one of the articles Amar listed had the formula using

buffersize = db_block_size * multi_block_read_count
the other had
buffersize = db_block_size * db_file_direct_io_count

My multiblock_read_count is 64. With either formula it doesn't appear I am using as much large pool as it says I should.

Any ideas?

Jodie