Memory used by oracle 9i processes

[Dave Walley]

Hi. It has come to my attention that the oracle 9i processes are reserving a lot more memory than the equivalent oracle 8i processes were. For example the res mem value
for 8 and 9 are as follow

8i ora_smon 10m
9i ora_smon 46m

8i ora_pmon 10m
9i ora_pmon 46m

all other usual processes are equally large under 9i.

What dictates how much memory is reserved by each process?

Thanks for your help.

Dave

[Peter Godron]

Dave,
had a quick look in metatool and the info I found was:
Size is governed by parameters for db cache, shared pool and other memory "pools". Background processes (like ckpt, dbw0, lgwr, pmon, smon) map to this shared segment.
Regards

[Steven E. Protter]

init.ora effects memory use as well.

Also, this oracle is bigger. Not that this bothers me, its more reliable than its predessor as well.

SEP

[Stuart Abramson]

You are absolutely right; and as I recall the Oracle DBAs at my last job knew that this was going to happen and warned us. Also the disk space went up as well (we ran Oracle Applications). We also went from 32-bit to 64-bit, which has an impact, I think.

"ps -elf" shows you space that a process takes, and we wrote scripts to compare them. man ps will show you which field it is.

[TwoProc]

Dave,

There's one constant out there - every upgrade requires "more". More of - whatever you have the least of to spare... :-)

I too noticed great increases in memory required in 9i, as well as newer versions of Applications.

The good news is - if you feed 9i enough memory - I believe it has the ability to outrun your 8i database much more so than if you did the same (memory increases) for 8i.

Also, had you not made the jump to 64bit - you could never have stood up the larger SGA's that many larger databases require.

[Julio Yamawaki]

Hi,

What about db_block_size parameter?
If you use db_block_size, say, 1k in Oracle 8i and 8k in Oracle 9i, when you multiply db_block_buffers (this parameter was substituted by db_cache_size in Oracle 9i, but you can use this as a static parameter), by 8k, you will have a 8 times bigger buffer.
Check this!

Regards,

[Anand Rao]

Hi,

Another very important thing is the 64-bit addresses. the database is 64-bit and the OS is 64-bit.

This means 64-bit adress structures, pointers, links, variables, etc. So, the increase in memory consumption is basically expected due to the larger variables and structures.

Also, in ps -eaf or ps -aux commands, the SZ or SIZE column will include the total SGA size also. glance and top will report huge sizes for each process under this column since each process also attaches itself to the SGA.

a pstack or a pmap is better to accurately estimate the memory consumption of these background processes or a user shadow process.

regards
anand

[Dave Walley]

Thanks for you advice. Dave