L2000 tuning for higher perf

[Paprocki]

Hi all,
I have a L2000 HPUX 11.00 whith 3GB of RAM and 6GB of swap.
I have an Oracle database 8.0.6 with an ERP "generix"
The server is limited in perf with this environnement (200-250 processes for generix), could u guide me ?
Do i have to change kernel parameters..?
What else should i do..?

thanks!!

[Pete Randall]

More RAM would be the best possible thing you could do.


Pete

[Eric Antunes]

Hi,

Post the results of:

HP-UX Kernel:

#kmtune -l -q shmmax
#kmtune -l -q dbc_max_pct
#kmtune -l -q dbc_min_pct
#swapinfo -tam

Oracle init.ora parameters:

processes = ?
db_block_buffers = ?
db_files = ?
log_checkpoint_interval = ?
log_buffer = ?
shared_pool_size = ?
dbwr_io_slaves = ?
db_block_lru_latches = ?

Best Regards,

Eric Antunes

[Bharat Katkar]

Hi,
This doc may be helpful.
Regards,

[Paprocki]

here the results..

kmtune -l -q shmmax :
Parameter: shmmax
Value: 0X4000000
Default: 0X4000000
Minimum: -
Module: -

l2000-1:/#kmtune -l -q dbc_max_pct
Parameter: dbc_max_pct
Value: 10
Default: 50
Minimum: -
Module: -

l2000-1:/#kmtune -l -q dbc_min_pct
Parameter: dbc_min_pct
Value: 5
Default: 5
Minimum: -
Module: -
l2000-1:/#swapinfo -tam
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 6000 0 1024 0% 0 - 1 /dev/vg00/lvol2
reserve - 1024 -1024
memory 2404 2023 381 84%
total 8404 3047 381 36% - 0 -

oracle:

processes=794
db_block_buffers = 30000
db_files = 40
log_buffer = 32768
log_checkpoint_interval = 4194304
shared_pool_size = 30000000

and dbwr_io_slaves = ?
db_block_lru_latches = ?
are not present...

any idea..?

[Eric Antunes · Accepted solution]

Ok,

HP-UX:

You should increase shmmax to 2147483648 (You can do this on SAM->Kernel Configuration->Configurable Parameters->[select shmmax]->Actions->Modify Configurable Parameter...)

It seems you have to much swap since you are not using any of the 6Gb available. But this not affects performance, only disk space.

Oracle:

Try to decrease db_block_buffers (if you have db_block_size = 8192, then you are reserving aprox. 240Mb for buffer cache: that's too much!):

db_block_buffers = 10000

Increase log_buffer:

log_buffer = 1048576

dbwr_io_slaves should only be used for heavy OLTP processing databases.

dbwr_io_slaves should only be used id you have multiple processors. For exemple, if you have 3 processors, you should use:

dbwr_io_slaves = 3
db_block_lru_latches = 3


Best Regards,

Eric Antunes

[Eric Antunes]

Sorry, where I wrote:

"dbwr_io_slaves should only be used id you have multiple processors. For exemple, if you have 3 processors, you should use:

dbwr_io_slaves = 3
db_block_lru_latches = 3"

You should read:

"
db_writer_processes (I'm not sure) should only be used id you have multiple processors. For exemple, if you have 3 processors, you should use:

db_writer_processes = 3
db_block_lru_latches = 3
"

[Eric Antunes]

Hi,

Let me know if performance gets better.

Eric Antunes

[Paprocki]

ok and thxs..i'll tell u this monday.. ;-)