DB_block_buffer

Hi,
Can anybody tell me, how can i modify the db_block_buffer in init.ora file ?
I don't want to modify manually in the init.ora file.
So, which tool, i have to use ?
Thanks in advance.

James, (you failed to list your Oracle version if would have been helpful) If your database is using an init.ora file instead of an spfile which is recommended you edit the init.ora file with your favorite text editor.
With an spfile you use alter system set parameter = value scope=spfile|both
See the SQL manual.
HTH -- Mark D Powell --

Similar Messages

Oracle preformance

Hi,
I have problem with perfomance of my program
that import some data from binary file to
oracle.
When a set small db_block_buffers cca 40 (db_block_size = 16384) imprted time plus aroud operation is 5 min (file have about 14MB = cca 100000 rows) it is OK.
So, when I change db_block_buffers to the 100, 600 and other greater then 70 process, time rapidly increases to the 15 min. It is terrible for me.
So, I need for other proces bigger db_block_buffer.
RAM 2GB
shared memory 1GB
linux 7.1
SGA = 512 MB
datafile on raw disks on RAID5
What can I do ?
thanks
Martin

Maybe Oracle and Linux make buffering
You try set chattr na oracle files and lock
SGA (LOCK_SGA in init.ora don't work locking,
I make litle program in C for locking shared memory)
null

Oracle Instances Processes consuming too much memory

Will somebody be able to guide me with the factors that are causing too much memory consumption when a single Oracle.exe instance is started ? Due to this problem it is not possible to run 2 instances at the same time.
Oracle is running on Windows NT 4.0 SP6.
Oracle version 8.1.6.0.0
Currently we have 512MBytes of ram.(4.83G free space)
One of the Instance consumes as much as 178,320K of Ram,
another consumes about 138,000K

Oracle consumes as much memory as you want, but more memory is faster.
Review your init.ora (db_block_buffer y shared_pool_size)
(It4s posible to run 10 instances in a machine with just 1 Gigabyte and they work fine. It depends on the use of the databases, the number of users, just for production or development, etc, ...).

Oracle 8i giving slow response when running sql query

hi ,
i am using oracle 8i server ;
and sql query giving slow response before some days;
so which parameter need to increase
DB_BLOCK_BFFERS OR SHARE_POOL_SIZE
CURRENT SETING IS
DB_BLOCK_BUFFER=256000;
SHARE POOLSIZE=300M;
pls advice to me
thanks

No parameters need to be increased. Tuning is not done but making arbitrary changes to values and hoping that good things will happen by accident.
The thing you really should be worrying about is your antique database and likely its impact on your skill set unless you are one year away from retirement.
Where's the actual version number?
Where's the DDL?
Where's the DML?
Where's the Explain Plan?

Tuning of SGA on Oracle 8.0.5 and AIX 4.3.3

Hi folks,
I have an AIX box running AIX 4.3.3 and Oracle 8.0.5.2.0. The original SGA settings are as follows:
shared_pool_size = 150000000
sort_area_size = 1024876
db_block_buffer = 126102
The SGA size then is:
Total System Global Area 1838467232 bytes
Fixed Size 47264 bytes
Variable Size 803287040 bytes
Database Buffers 1033027584 bytes
Redo Buffers 2105344 bytes
I changed the SGA settings to the following:
shared_pool_size = 768000000
sort_area_size = 512000000
The SGA size is now:
Total System Global Area 2227943584 bytes
Fixed Size 47264 bytes
Variable Size 214097920 bytes
Database Buffers 2013265920 bytes
Redo Buffers 532480 bytes
After this change was done, the system ran but the users had some problems running reports in JDEOneWorld.
I then downsize the SGA settings and altered the db_block_buffer to improve buffer cache hits with these settings:
db_block_buffers = 245760
shared_pool_size = 200000000
sort_area_size = 1048576
On restarting the database with the new values, I noted that the SGA size remained unchanged as below:
Total System Global Area 2278475936 bytes
Fixed Size 47264 bytes
Variable Size 264105984 bytes
Database Buffers 2013265920 bytes
Redo Buffers 1056768 bytes
I tried to increase the size of db_block_buffers to 327680 but encountered ORA-27123 when trying to startup the database. I did some searches in metalink and found an article saying that the SGA size cannot exceed 2GB on AIX platforms. (Doc ID: 114934.1). Therefore I resize the db_block_buffers back to 245760.
What puzzles me is why the SGA size does not reflect my latest init.ora settings? Does anyone have any idea on how I can get the SGA to reflect the updated values in the init.ora file?

I did a find and here are the files I got:
278529 12 -rw-r--r-- 1 oracle dba 8332 May 12 1998 /u01/app/oracle/product/805/dbs/init.ora
278531 0 lrwxrwxrwx 1 oracle dba 43 Apr 14 2001 /u01/app/oracle/product/805/dbs/initklb.ora -> /u01/app/oracle/admin/klb/pfile/initklb.ora
278532 0 lrwxrwxrwx 1 oracle dba 45 Apr 14 2001 /u01/app/oracle/product/805/dbs/initklb_0.ora -> /u01/app/oracle/admin/klb/pfile/initklb_0.ora
49154 8 -rw-r--r-- 1 oracle dba 4657 Sep 7 2000 /u01/app/oracle/product/orig_db/initklb.ora
49155 8 -rw-r--r-- 1 oracle dba 4657 Jul 27 2000 /u01/app/oracle/product/orig_db/initklb_0.ora
151553 4 -rw-r--r-- 1 oracle dba 2985 Aug 4 2006 /u01/app/oracle/product/luz/pfile/initklb.ora
49161 8 -rw-r--r-- 1 oracle dba 4118 Oct 11 11:22 /u01/app/oracle/admin/klb/pfile/initklb.ora
135169 4 -rw-r--r-- 1 oracle dba 2896 Aug 4 2006 /u01/app/oracle/admin/klb/pfile/backup/initklb.ora
49168 8 -rw-r--r-- 1 root dba 4687 Jul 21 2001 /u01/app/oracle/admin/klb/pfile/archive/initklb1.ora
49162 8 -rw-r--r-- 1 oracle dba 4657 Dec 23 1999 /u01/app/oracle/admin/klb/pfile/archive/initklb_0.ora
The instance name is KLB so the applicable files are:
278531 0 lrwxrwxrwx 1 oracle dba 43 Apr 14 2001 /u01/app/oracle/product/805/dbs/initklb.ora -> /u01/app/oracle/admin/klb/pfile/initklb.ora
278532 0 lrwxrwxrwx 1 oracle dba 45 Apr 14 2001 /u01/app/oracle/product/805/dbs/initklb_0.ora -> /u01/app/oracle/admin/klb/pfile/initklb_0.ora
49154 8 -rw-r--r-- 1 oracle dba 4657 Sep 7 2000 /u01/app/oracle/product/orig_db/initklb.ora
49155 8 -rw-r--r-- 1 oracle dba 4657 Jul 27 2000 /u01/app/oracle/product/orig_db/initklb_0.ora
151553 4 -rw-r--r-- 1 oracle dba 2985 Aug 4 2006 /u01/app/oracle/product/luz/pfile/initklb.ora
49161 8 -rw-r--r-- 1 oracle dba 4118 Oct 11 11:22 /u01/app/oracle/admin/klb/pfile/initklb.ora
135169 4 -rw-r--r-- 1 oracle dba 2896 Aug 4 2006 /u01/app/oracle/admin/klb/pfile/backup/initklb.ora
49168 8 -rw-r--r-- 1 root dba 4687 Jul 21 2001 /u01/app/oracle/admin/klb/pfile/archive/initklb1.ora
49162 8 -rw-r--r-- 1 oracle dba 4657 Dec 23 1999 /u01/app/oracle/admin/klb/pfile/archive/initklb_0.ora
Now, exclude the files in the product, backup and archive directories (the pfile should be in the pfile directory), and that leaves us with:
49161 8 -rw-r--r-- 1 oracle dba 4118 Oct 11 11:22 /u01/app/oracle/admin/klb/pfile/initklb.ora
(Which is the file that I edited)
As I mentioned previously, my client is not prepared to perform an upgrade on both Oracle and AIX. As much as I would like to have them upgrade to the latest versions, it is just not possible.
Oracle 8.0.5 does not have spfiles, so there is no way for me to edit a spfile unless this is one of those undocumented tricks floating around.

Windows 32bit OS 에서 Oracle Memory Sizing

제품 : Database
작성날짜 : 2007-12-21
PURPOSE
Windows 32bit OS 메모리 제약과 Oracle 에서 추가메모리 사용하도록 설정하는 방법을 설명한다.
CONTENTS
1. Windows 32bit 메모리 확장 관련 용어들
2. Windows 32bit 메모리 확장 설정 방법
3. AWE_WINDOW_MEMORY 파라미터 산정
4. Windows 에서 Oracle 메모리 사용량 모니터링
5. Windows 에서 Oracle 메모리 사용량 줄이는 방법
EXPLANATIONS
1. Windows 32bit 메모리 확장 관련 용어
- /3GB
Windows 32bit OS 에서 메모리 어드레싱은 4GB 까지 가능하고, 이 중 2GB 는 커널이 사용하고
나머지 2GB 가 사용자 프로세스에게 할당되는 최대 메모리 크기이다. Windows 2000 이상에서는
/3GB 옵션을 사용하여 커널에 1GB 만 할당하고 나머지 3GB 를 사용자 프로세스에게 할당하여
2GB 한계를 넘어 메모리를 확장하여 사용할 수 있다. OS 에서는 boot.ini에 /3GB 옵션 추가하고,
어플리케이션 실행파일 헤더에 IMAGE_FILE_LARGE_ADDRESS_AWARE bit 이 설정되어 있으면 된다.
Oracle 8.1.7 이상에서는 기본적으로 IMAGE_FILE_LARGE_ADDRESS_AWARE bit 설정되었으므로 3GB
사용가능하다. 참고로 Windows 64bit OS 는 한 프로세스당 어드레싱 가능한 메모리가 8TB 이므로
32bit 에서 사용하는 추가적인 옵션들이 필요없다.
- /PAE (Physical Address Extentions)
/3GB 옵션과 더불어 32bit 메모리 제약을 넘기위한 방법으로, 가상 메모리내에 포인터 테이블로
"별도 메모리 영역"을 매개함으로 프로세스가 사용할 수 있는 실제 메모리 크기를 확장한다.
예를들어 Oracle 은 가상 메모리 2GB 영역 (/3GB 사용하면 3GB) 내에 포인터 테이블(window)만
저장하고 실제 DB BUFFER CACHE 는 "별도 메모리 영역" 에 할당한 후, BUFFER CACHE 사용시
포인터 테이블을 매개하여 실제 BUFFER CACHE 영역 사용한다. 여기서 "별도 메모리 영역" 도
메모리를 사용하는 것이다.
- AWE (Address Windowing Extensions) & Oracle 구현방식
AWE 는 PAE 기능을 사용하도록 어플리케이션에 제공된 API 이다.
AWE API 를 사용한 오라클 버젼들은 아래와 같다.
* Oracle 8.1.6.x
* Oracle 8.1.7.x
* Oracle 9.2.x
* Oracle 10.1.x
* Oracle 10.2.x
주의! 9.1.0.x 는 AWE 지원안됨.
AWE 구현된 Oracle 버젼들은 3GB 이상의 SGA 사용할 수 있는데, 정확히 말하면 SGA 중 BUFFER CACHE
영역만 3GB 이상의 "별도 메모리 영역"에 지정할 수 있다. 오라클에서 AWE 를 사용하면 BUFFER CACHE
사용시 포인터 테이블과 실제 BUFFER CACHE 가 있는 별도 메모리 영역과의 mapping & unmapping
과정이 추가되므로 기본적인 처리보다 느릴 수 있지만, 적어도 IO 을 줄일 수 있기 때문에
IO 느린 시스템에서는 성능 향상을 기대할 수 있다.
2. Windows 32bit 메모리 확장 설정 방법
- boot.ini 에 /3GB /PAE option 을 추가
multi(0)disk(0)rdisk(0)partition(1)\WINNT="Microsoft Windows 2000 Advanced Server" /3GB /PAE
- 오라클 서비스 실행하는 OS USER 에 "Lock memory pages"의 권한 부여
. Start -> Programs -> Administrative Tools -> Local Security Policy
(on a Domain Controller, click on 'Domain Security Policy' instead of 'Local Security Policy')
. Double-click on the 'Lock Pages in memory' policy.
. Add the appropriate user and click 'Ok'.
. Restart the OracleService<SID>
- 포인터 테이블 (window) 크기 지정하는 AWE_WINDOW_MEMORY 를 레지스트리 추가
. HKEY_LOCAL_MACHINE => Software => Oracle => HomeX 에 AWE_WINDOW_MEMORY=1024000000 추가
주의! 여러 인스턴스라면 각 Home 에 추가
- init.ora 에 USE_INDIRECT_DATA_BUFFERS=TRUE 추가하고 DB_BLOCK_BUFFERS 로 BUFFER CACHE 크기 설정
주의! DB_CACHE_SIZE 로 BUFFER CACHE 지정하면 에러 발생한다.
ORA-00385: cannot enable Very Large Memory with new buffer cache parameters
- Oracle10g 에서는 SGA_TARGET=0 설정하여 Automatic Memory Management 를 disable 해야한다.
3. AWE_WINDOW_MEMORY 크기 계산 방법
(1) 기본값 : AWE_WINDOW_MEMORY = 1GB
/PAE /3GB 를 지정해도 아래 메모리들의 총합은 3GB 내에 들어와야한다.
- AWE_WINDOW_MEMORY
- BUFFER CACHE 제외한 SGA (shared_pool, large_pool, java_pool, log_buffers ...)
- Overhead for Oracle.exe DLLs (100MB 정도)
- Stack space per thread (1MB/thread)
- PGA and UGA
결국 BUFFER CACHE 최대 크기는 OS 메모리 - 4GB + AWE_WINDOW_MEMORY 로 산정된다.
(여기서 4GB는 Oracle이 사용하는 3GB와 OS에서 사용하는 1GB의 합. 결국 이 영역을 제외한 나머지 영역에 buffer cache를 지정하여 사용할 수 있다.)
예를들어 12GB RAM 시스템에서 최대 BUFFER CACHE 크기= (Total RAM - 4GB + AWE_WINDOW_MEMORY) = 12GB - 4GB + 1GB = 9GB 이다.
(2) 8.1.7 이하에서는 AWE_WINDOW_MEMORY 최소값 제한이 없었으나,
Oracle9i R2 부터 최소값보다 작게 설정하면 DB 기동시 에러들이 발생한다.
ORA-27102 out of memory
OSD-00034 Message 34 not found; Product=RDBMS;facility =SOSD
O/S Error: (OS 8) Not enough storage is available to process this command
AWE_WINDOW_MEMORY 최소값 계산 공식
a. MIN(AWE_WINDOW_MEMORY)= (4096 * DB_BLOCK_SIZE * DBBLOCK_LRU_LATCHES)/8
b. DBBLOCK_LRU_LATCHES = (Max buffer pools * SETS_PER_POOL)
c. Max Buffer Pools = 8 (상수)
d. SETS_PER_POOL /* USE_INDIRECT_DATA_BUFFERS=TRUE 로 VLM ENABLED */
SETS_PER_POOL = 2*CPU_COUNT (if VLM is enabled)
SETS_PER_POOL = CPU Count/2 (if VLM is NOT enabled)
시나리오 #1:
# of CPU's = 8
DB_BLOCK_SIZE = 8192
Total RAM = 8GB
SETS_PER_POOL = 2 * CPU_COUNT = 16
DBBLOCK_LRU_LATCHES = (max buffer Pools * sets_per_pool) = 8*16 = 128
MIN(AWE_WINDOW_MEMORY) = (4096*DB_BLOCK_SIZE * DBBLOCK_LRU_LATCHES) / 8
= (4096 * 8192 * 128) / 8 = 536870912 bytes = 512 MB
시나리오 #2:
# of CPU's = 16
DB_BLOCK_SIZE = 8192
Total RAM = 16 GB
SETS_PER_POOL = 2 * CPU_COUNT = 32
DBBLOCK_LRU_LATCHES = (max buffer Pools * sets_per_pool) = 8*32 = 256
MIN(AWE_WINDOW_MEMORY) = (4096*DB_BLOCK_SIZE * DBBLOCK_LRU_LATCHES) / 8
= ( 4096 * 8192 * 256) / 8 = 1073741824 bytes = 1024 MB
주의 !!!
1. ORA-27102 는 AWE_WINDOW_MEMORY 최소값 이하일때도 발생하지만, BUFFER CACHE 가
실제 메모리보다 크게 지정된 경우에도 발생한다.
2. ORA-27102 는 DB_BLOCK_BUFFER < AWE_WINDOW_MEMORY 일때도 발생한다.
3. AWE_WINDOW_MEMORY 설정값은 "계산된 최소값 + 10 MB (overhead)" 으로 설정해야한다.
4. Hyperthreading enabled 시스템의 경우 # of CPU's 를 두배로 계산해야한다.
4. Windows 에서 Oracle 메모리 사용량 모니터링
TASK MANAGER 에서 VM 값은 정확하지 않으므로 관리도구에서 별도 모니터링해야한다.
관리도구 => Perfomance 선택 => "추가" => 성능개체:Process, Count:Virtual Bytes, Instance:oracle.exe 선택
5. Windows 에서 Oracle 메모리 사용량 줄이는 방법
(1) Oracle 실행파일 스택 크기 조정
Thread 들이 미리 1MB 스택을 PGA 로 선점하고 필요하게 되면 더 확장해서 사용하는데
대부분들의 Thread 들은 초기 스택도 사용하지 않는 경우가 많다.
아래처럼 default=1 MB 를 500 KB로 줄일 수 있다.
. DB 내린 후 ORACLE_HOME/bin 에서 수행
C:\OH\bin> orastack oracle.exe 500000
C:\OH\bin> orastack tnslsnr.exe 500000
C:\OH\bin> orastack svrmgrl.exe 500000
(2) Shared Server (pre-Oracle9i 는 MTS 설정) 사용: 세션들이 메모리를 공유하므로 Oracle 프로세스 전체적인 메모리를 줄일 수 있다.
References
Note 225349.1 - Implementing Address Windowing Extensions (AWE) or VLM on Windows Platforms
Note.205089.1 - Oracle and Hyperthreading

Basically ORA-12154 means, service name can't be found in tnsnames.ora or this file can't be located at all. Are there multiple ORACLE_HOMEs on the machine ? Is the application based on OLEDB running in a different home , NOT in the client's ORACLE_HOME? You may set TNS_ADMIN, pointing to your tnsnames.ora, as global environment variable. So each application on the machine will know this location.
I'm sure this is not a problem 64bit versus 32bit.
Werner

New Database parameter settings

I have a new Windows server with Oracle 11g, Windows 2008 R2, 4 CPU and 10 G of memory.
I had another server that had 2 cpu and 5 G of memory but the Oracle database was not utilizing the memory appropriately (I am also New to Oracle)
So I am setting up the new database to use 8G for the SGA and PGA this time on this new server. One thing I noticed in the last test of the application on the previos server was that the bottleneck appeared to be the Buffer cache hit ratio and shared pool....it also appeared to be extremely IO intesive however the network adapeter never spiked above 12%. So all of the contention was or seemed to be in the database.
With the new server should I increase the db_block_buffer parameter? It is currently defaulted to 0?
Any help would be appreciated. The application we are setting this up for is extremely cursor intensive and we cannot change the code.

ofmdba wrote:I have a new Windows server with Oracle 11g, Windows 2008 R2, 4 CPU and 10 G of memory.
I had another server that had 2 cpu and 5 G of memory but the Oracle database was not utilizing the memory appropriately (I am also New to Oracle)
>
How did you measure that oracle is not using the memory properly? Oracle is not supposed to use the complete physical memory anyways. It can use only what is going to be set for it.
<font color="red"> I was watching spotlight for Oracle and the server perfmon counters and never saw any real work....Physical IO appeared high the total SGA area was 888MB and the appliation that goes against it is cursor intensive....at the time a tester was running a search and sort of 14,000 records and it took 18 minutes. My thought, however wrong I might be, was that the SGA could be sized much larger to take advantage of the memory on this server, and hopefully reduce the time it is taking </font>
>
So I am setting up the new database to use 8G for the SGA and PGA this time on this new server. One thing I noticed in the last test of the application on the previos server was that the bottleneck appeared to be the Buffer cache hit ratio and shared pool....it also appeared to be extremely IO intesive however the network adapeter never spiked above 12%. So all of the contention was or seemed to be in the database.
>
Using the hit ratios to tell that there was a contention or bad performance is not such a good idea. If hit ratio was the only criteria that you used or had , I wouldn't call it a right move to make.
>
With the new server should I increase the db_block_buffer parameter? It is currently defaulted to 0?
>
This parameter is obsoleted and not in the current release of 11g but since 9i. So you should not bother about this parameter at all.
>
Any help would be appreciated. The application we are setting this up for is extremely cursor intensive and we cannot change the code.
>
So if its creating lots of cursors , it is possible that the application is using lots of literals and constants which can be reduced using the parameter Cursor_sharing.

I can't set USE_INDIRECT_DATA_BUFFERS.

I tried to set use_indirect_data_buffers is true and I changed sga_target=0 ,other cache_size parameter = 0, db_block_buffer =0 and db_block_size=8192.
When I start database will show this error
ORA-00371: not enough shared pool memory, should be atleast xxxxxx bytes
I think I should set value to db_block_buffer parameter but I have no idea what value to it.
Please advise.

Dear all,
I know it a very old thread but i am opening it because it related to VLM.I have some doubts.
ENV:10.2.3 database running on redhat linux el4 with kernel 2.6.9-42.ELsmp.
1)We are not able to give memory more than 3 GB to oracle i mean SGA in order to give more memory we decided to implement VLM.I want to know when we enable VLM we implement ramfs and we have tempfs currently runnign so ramfs cannot be run with tempfs so we need to disable tempfs.
2)We have 8gb on db server we want to give atleast 5gb to SGA..in this scenario VLM will help us?
3)if i enable ramfs swap will get used as i see whne i implement VLM swap usage is 0
Regards
Edited by: hungry_dba on Feb 23, 2011 9:22 AM

Oracle 8i shared memory

Hi every one,
i have problem installing oracle 8i database in my branded system hp-PIV system. while it gives the following error during initilizing database
ORA 04031: unable to allocate 'size in number' of shared memory ("shared pool" ,"unknown object ", "sga heap"," db_block_buffer")
while it installs fine in all hp-PIII machines but gives above error in all hp-PIV branded machines.
oracle 9i installs accurately in all machines. what is the actual problem?
Regards & Thanks
Najeeb

04031, 00000, "unable to allocate %s bytes of shared memory (\"%s\",\"%s\",\"%s\",\"%s\")"
// *Cause: More shared memory is needed than was allocated in the shared
//          pool.
// *Action: If the shared pool is out of memory, either use the
//          dbms_shared_pool package to pin large packages,
//          reduce your use of shared memory, or increase the amount of
//          available shared memory by increasing the value of the
//          INIT.ORA parameters "shared_pool_reserved_size" and
//          "shared_pool_size".
//          If the large pool is out of memory, increase the INIT.ORA
//          parameter "large_pool_size".

SGA db_buffer_cache

Hi,
We are in process of tuning the SGA size of oracle database 8.1.7.0
We want to increase the size of db_buffer_cache.
So, which other dependent parameters we have to take care of while increasing the size of db_buffer_cache ?
With Regards

Hi,
Our database in on Windows OS.
But what ratio we have to keep in mind so that we dont get into page swapping.?
We are having total SGA size of 180 MB and database bufffer size of 53 MB
We are also having sufficient RAM to increase the size of SGA.
db_block_buffer size is 6847
db_block_size is 8192
database buffer size is 56090624 bytes
so what is the suggested size can we have for our db_block_buffer ?
With Regards

Cannot set sga_target with db_block_buffers set

Hello Oracle Community!
I'm trying to set a sga_target on my Oracle 10g R2 database but it shows me the following message: Cannot set sga_target with db_block_buffers set
What exactly's the impact on disabling db_block_buffers parameter? Is it better to do while the database's not being used?
Thank you very much!

K- wrote:
@Aman
This database uses SPFILE, how do I proceed to erase the db_block_buffer on it? The only way's using the CREATE PFILE FROM SPFILE command? Do I have to shutdown my production database if I execute it?
ThanksYes, this is a major change, you will have to bounce. Probably the way to go would be to create pfile from spfile, copy that away somewhere safe, edit the pfile, shutdown and startup with the pfile. If you've made a mistake at this point where it won't come up, you can startup without the pfile if you need to, or keep trying to fix the pfile. When you are up, check that the parameters are what you think you did, then create spfile from pfile and bounce. Of course, if performance then goes wacko, you can go back to the saved pfile until you figure out what went wrong.
You can try with the sizing you are currently using, then after running for a while use the advisors to tweak. Note that with sga_max_size it can vary by platform and configuration whether you use the difference between target and max, so platform determines whether you should make them the same (see the thread sga_max_size vs sga_target ) . If they should be the same, you'll have to bounce later when you tweak. Get familiar with the HIST views that record the self-tuning changes as they happen.

": Not enough physical memory for SHM_SHARE_MMU "

Dear ALL,
": Not enough physical memory for SHM_SHARE_MMU "
This is an error message which occurs while trying to mount the database on startup. I've two database running of solaris. DB1 is running with 700M shared pool size. I want to run the DB2 database with 400M. But if i specify 400M shared_pool size on DB2 it is not mounting. It is only mounting when I specify 200M size of shared_pool. Machine has 4cpu and 4gb of RAM.
I've configured the kernel and shemaphore parameter according to the oracle process.
Any thoughts or suggestions appreciated.

The whole SGA must fit into physical memory (SGA=shared_pool_size + (db_block_buffer*db_block_size) + log_buffer + java_pool_size).

Why backup (dmp) restore at change in init.ora

hello,
i had change db_block_buffer and share_pool_size increase more in size
due to time saving at report retrive/process.
should i remove all tables and restore all table again.
thanks
pankaj raval

pankaj raval wrote:
hello aman,
i want to increasing processing speed/ time for report retrieving.
so i should be increase the of SGA what i had asked previously in this forum.
but some ask me to bounce the database in this regarding.
and i had change SGA parameter what i post in before, and check process time
and it gives me proper result , so i want to know why some DBA asked me
to bounce database again?
rega.
pankaj ravalShould I use some strong words for your dba? Well, let me say this that increasing the memory and expecting the queries to be faster is rather a wrong approach to tune the queries. If you are saying that doing so gave you proper results you are sadly mistaken because when the memory would be full, you would again see the performance coming down and I can bet, you won't have the luxry to add memory as and when you would want. Second, you were asked what version of database you are running because the parameter, db_block_buffer has been deprecated since 9i onwards and you must be using db_cache_size . Depending on the version of the db(which you haven't shared with us) , most of the memory regarding parameters are now auto-tuned thus not requiring you to do anything about them and are also dynamic thus doesn't need to bounce the db as well.
HTH
Aman....

V$latch_misses - latches

I start to read the note 22908. I don't understand a phrase about the buffer cache latches. The phrase is
Before of incrementing the parameter db_block_buffer check that specific blocks are not causing the contention avoiding memory wasting.
Anyone can help me to understand this issue with an example or something ?
Thanks to all
Walter

The buffer cache is a memory structure that holds database blocks that have been requested to satisfy some SQL statement. Each time Oracle needs to read one of these blocks in memory, it needs to acquire a latch (a type of lock) to prevent another process modifying the block while it is being read. Whenever it tries to acquire a latch and cannot, the latch miss count is incremented.
If the buffer cache is too small, this can happen frequently, and the solution is to increase the size of the buffer cache by increasing the paramter db_blcok_buffers. However, you can also see a high number of latch misses when you have some "hot" blocks. A hot block is one that is very frequently accessed by multiple processes.
Since only one process can latch a block at a time, this causes a higher nmber of latch misses. One of the more common causes of this type of activity is when you have a small lookup table that fits into one or two blocks. If this table is accessed frequently, then it will always be in buffer cache, and since everyone is trying to access it, it will frequently be latched, causing misses. Since most of the latch misses are being caused by attempts to latch only these few blocks, then increasing the size of the buffer cache will not solve the problem.
HTH
John

Data Movement into a new DB via dblink - Consideration

Hi,
I am looking for some tips to consider before creating my pfile & DB. ( in terms of the MAIN memory parameters and the like )
I know this is a rather crazy thing to ask considering the zillions of params, but here is my situation :
This DB will be used to archive old data from a 7.3.4 to a another 7.3.4 DB on a solaris box. The data will be moved via a db link.
Do I go in for a bigger db_block_buffer & log buffer ; 5 - 6 log groups of 30 MB ? Would a rollback segment sizing of 5mb (initial, next) be sufficient if I create a total of 10 such segments ?
Fact of the matter is we are not sure how much amount of data would be moved to the new database from the existing one. It depends on some date factor which considers the archiving capacity.
The memory on the archiving box is 512MB and already has 5 other such archive databases running on it.
Could anyone please suggest a right way when considering the new parameters ?
Thanks.

Here are some thoughts:
- First ensure you have a recovery plan in case this goes wrong.
- It some as though you are expecting to do a few large transactions which are not concurrent. For this a small number of large rollback segments is probably better.... perhaps even as low as 3. Ensure all have a good minimum size and try to avoid shrinking. In my view 3 x 100mb would be better than 10 * 30mb. For the actual size required that is your to work out.
- I have a very rough rule of thumb that the size of the sum ot the SGA's should be between 1/3 and 2/3 of main physical memory unless there's a good reason otherwise (you actually need to factor in how heavily databases used, number of connections, size of operating system, what else is on the server), but the rule has proved a a good starting point over the years. With Solaris Oracle 7.3.4 you might need to be a little careful at the top end of that as I seem to remember SGA's are locked in memory at 7.3.4 (not for some later versions though), and I once experienced once a horrible thrash with solaris 7 ( I dont think Solaris 8 is as bad in this situation).
- For your data take on you may be able to shut down some databases at that may help. You may also be able to create a bigger db_block_buffer temporarily and that may help.
- From memory uping SORT_AREA_SIZE and SORT_RETAINED_AREA_SIZE may help as you should have only a few connections.
- Ensuring you are using a IPC sqlnet connection may help.
- You will be lucky if you you can do anything to easy i/o contention, but if you can it will help.
- Ensure your redo logs are adequate sized. Be aware if you will be doing more work updating and inserting relative to querying ... in that case pay more attention to ensuing the redo logs are not your bottleneck.
You of course take resonsibility for any decisions you may take. I have not taken a long time in these thoughts or checked them for completeness - rgds bigdelboy
Depending on the support status of your server you may be able to add more ram to it cheaply or otherwise.

DB_block_buffer

Similar Messages

Maybe you are looking for