Definition of size of SGA in 10g

Hi all

We have a dual processor xeon server with 32 GB of RAM

We have Six live database

One of a many users and most used databases has about 150 users.

According to the Advisor of the memory, we have changed the size of the sga - sga_target from 1 G to 3 G,

After a few days, he advises again for 6 GB, now we put the sga_target 6 GB.

But now again the conclusions of the consultant/ADDM memory shows always SGA insufficiently size and needs to be increased to 10 GB.

What we should in this circumstance? We should raise them, or to look for something to improve?

Please notify.

Thank you

Vijayaraghavan.K

Doubling of the size of the SGA is estimated only to give you a 5% increase. Note the word estimated.

While it cannot be said with certainty that it can help performance significantly, which doesn't seem much of a gain, and may indicate that you need to look elsewhere performance improvements (if improvements are what you're really after). But if you have more memory on the server that you know what to do with, you can decide otherwise.

Tags: Database

Similar Questions

size of SGA in 10g r2
Dear Oracle gurus

Oracle 10.2.0.3.0 version (64-bit)

I use 2 cases on
SUSE Linux Enterprise Server 11 (with 24 GB memory)
(1) first: -.
SGA_MAX_SIZE 4016M
SGA_TARGET 4016M

pga_aggregate_target 1595M

(2) the second instance: -.

SGA_MAX_SIZE 1536M
SGA_TARGET 1536M
pga_aggregate_target 1595M

I intend to increase the sga to 8 GB of first instance.

All boards
Thanks for your time.
880674 wrote:
The Oracle gurus,
Thank you all for the reply.
Help, please.
I really want to learn the performance tuning, but I starter, so my basic are not that great.
Please suggest my good book for performance optimization.

As Pavan said, start with the manual of the Performance. Who has the basics. Then you should check book optimizer Jonathan Lewis to understand how to work on what is optimizer, Cary Milsap book to find out what is really wrong and by Antognini Oracle performance troubleshooting. Note that most performance issues are not magically fixed playing with CMS or other parameters of the instance. You are much better let the instance parameters where they are and check what is actually slow you down. There are many variables involved, and sometimes fixing a bottleneck can lead to a worse bottleneck.

>

Let me explain my problem again so that I can get your attention or assistance
I have 1 job that takes 5 hours on production. I cloned db again and fast box (24 GB of memory), same work completed with in 40 min. I thought that growing
SGA will give us more performance gain, so I tried to increase the sga to 12 GB to 4 GB.
I faced the error ORA-27102. So I followed ID 301830.1
and set kernel.shmall = 4194304
kernel.shmmax = 12884901888
When I changed the sysctl.conf file and changed the shmall shmmax allowing 12 gig Oracle sga, increased runtime to 15 min; I don't know what is originally this.feel like shooting myself.

No need to shoot yourself, you just don't need to use the available tools to find out what the problem is. There are various things you can look at and tweak, but according to a methodology as Cary you can get right to the problem at hand. Sometimes even the statspack can point to the real problem.

It could be that you run in waste memory because you are or that you do not use hugepages, do not allow enough PGA for users to do their thing, who knows?

Instrumentation knows.

>

Please help/tips

SQL > startup strength;
ORA-27102: out of memory
Linux-x86_64 error: 28: no space is available on the device
SQL > startup
ORA-27102: out of memory
Linux-x86_64 error: 28: no space is available on the device

This should give you a clue what 'more' does not necessarily mean 'best '.

Size of SGA and PGA
Hi all

I have a question.
What are the best criteria to define the size of the SGA and PGA size in oracle 10g? and size in oracle 11g memory_target?
Is related to the size of the data? What is the criteria and factors which affects setting the size of the SGA, PGA (10 g) and memory_target (11 g)?

Kind regards
Hello

How can identify the best value for V memory_target $ MEMORY_TARGET_ADVICE?

We can take the example below:
```
SQL> select * from v$memory_target_advice;

MEMORY_SIZE MEMORY_SIZE_FACTOR ESTD_DB_TIME ESTD_DB_TIME_FACTOR    VERSION
----------- ------------------ ------------ ------------------- ----------
        393                ,75        17909              1,0004          0
        524                  1        17902                   1          0
        655               1,25        17900               ,9999          0
        786                1,5        17897               ,9997          0
        917               1,75        17897               ,9997          0
       1048                  2        17897               ,9997          0
```
You can see that 786 MB ESTD_DB_TIME_FACTOR beyond cannot be improved (it stuck here to 0.9997). So I may be updated (in this example) the MEMORY_TARGET 786 MB, I won't have any advantage to further increase.

Hope this helps.
Best regards
Jean Valentine

What should be the ideal size of SGA components?
Hi master,

We are using oracle 10G R2 on RHEL 4.7, I would like to know what should be the ideal size for components of the SGA, if my database size is 30 to 70 GB?

We need to set manually? or if we fix SGA_TARGET then it's OK?

Once again how much % should we specify SGA_TARGET if we * 8 Server CPU * with * 8 GB * of RAM?

and of course the database size will increase in function close...

any suggestion for it

Thank you and best regards

VD
If you want to manually set the values for pools, then his new return to the same answer that is trial and error.

For a well-written application (one that uses bind variables and has a minimal amount of literal SQL), a division of 40-60 of SGA should be well that is in the shared pool and other components 40-60% for the buffer cache, but I must add, YMMV.

size of SGA on hugemem kernel
Hello

I am running oracle 10.2.0.4 on RHELAS U4.6 and the kernel is Linux 2.6.9 - 67.ELhugemem #1 SMP i686 i686 i386 GNU/Linux.

Course of the system's RAM is 32 GB and the current size of the sga is 3G.

My question here is that it is possible that I can increase the size of the sga up to 8 GB?

Thank you for your cooperation.

Kind regards

Adnan Hamdus Salam.
Salvation;

Please see:
How to configure RHEL/OEL 4 32 bits for very large memory with ramfs and HugePages [317141.1 ID]< step="">
Linux HugePages: what it is... and what is not... [361323.1 ID]

Respect of
HELIOS

Size Max SGA in 32 bit OS
Hello

I have an Instance of Production EBS (12.1.1) in a 32-bit operating system (RHEL4.0).

RAM: 16 GB, SWAP: 8 GB

Our sga size is 1.7 GB

Is it possible to increase the size of our LMS.

Please advice.

Thanks in advance.
If your system is windows based on the 1.8 GB is the largest LVO for your database
If your Linux-based system the limit of physical memory max on a system 32-bit thusly imposed by no longer support the hugemem kernel is 16 GB :)

IF database-> RAC database: estimate of the size of the SGA for each instance

Hello!
Let's say that there is a single instance with SGA 100 G size DB (assume this size is optimal for this database).
If this database is intended to be converted to a RAC database with 2 instances (suppose DB workload will be spread over multiple instances).
How to estimate the needs of SGA size for each instance in the RAC database?
It takes 50G + General cool CARS (say 15%) for each instance?
(Do not consider situation when one of the instances has failed, and all the charge redirects to an instance having survived)
I'd be grateful if someone has offered or provided useful references.
---
Kind regards
Kyrylo

As a general rule, I would size the SGA on new nodes to 120% of the SGA to single instance. Yes, this means that you will more than double the size of the SGA. You will need additional space to hold the load memory Cache Fusion. And the buffer Cache are possibly a little more space to store additional versions of blocks that seem to accompany the many implementations of CARS. Now, it's just a rule of thumb, so it has a high degree to be very wrong.

You'll want to consider two big things... (1) the service being used. Will you take advantage of the services to perform the partitioning application? If so, instances can see various workloads that can lead to requests for different resources. (2) you will support a kind of failover (TAF or manual failover) applications? If so, then the LMS must be sufficiently sized to support the additional workload.

See you soon,.
Brian

Cloning of SGA size post

Hi all
I cloned 11.5.10.2 ebs however after cloning, I discovered sga size does not match the size of sga prod.
How could I change size to prod sga sga.
I use rhel 5.2 with ebs 11.5.10.2
Thank you

Hello

Please confirm that you are using a 64 bit OS.

If you do, then edit the file /etc/sysctl.conf have entered:

kernel.shmmax = 4294967296

Save and exit, then run the command 'sysctl Pei' and try to restart the DB.

If this does not work, please do:

1. ensure that RAM upgrade is detected by the program: free - m.

2 attach the output of the following command after the DB of the supply environment file:

Cat $ORACLE_HOME/dbs/init$ORACLE_SID.ora | grep sga_target

Thank you &

Best regards

SGA size limits
Hi all

Is there a list of notes or oracle metalink the limits of size of SGA on different platforms and 32/64-bit?
http://download.Oracle.com/docs/CD/E11882_01/server.112/e17110/limits004.htm#i288165

Creation of the database - the size of the SGA
Hi all

When I try to create a database using DBCA, my default SGA size: 128. PGA:794. When I click on the next button, I get the following caveat...

The following settings do not meet the requirements of the minimum recommended size:

A size of SGA to at least 160 MB is recommended

Do you want to Setup Wizard to change the setting of the database? I gave ignore, and I changed the size of the SGA at: 160, I received the warning below

The SGA size cannot be greater than the segment size memory maxmimum shared (1342177728). Refer to the guide to installing oracle for configrue your operating system kernel parameters.

Please suggest me on the same...

Thank you
RIF
user10737570 wrote:
My other SGA_MAX_SIZE database is 1 G.

Now, I am creating the database using DBCA. My default allocated sga size is 128 MB. Recommended size of SGA 160MO. I'm not able to extend the size of the SGA. Please refer my previous post...

Could you please tell me why my default screen is only showing 128 MB.

Thank you

The oracle Installer determines the sga_size bij available calculation free "shared memory".
It is the setting of the maximum shared memory such as published in the first message less memory used by your first database, less some other stuff shared memory.

To remedy, you increase the size of memory amount shared in your kernel and reboot the system if necessary.

If your system is a Redhat Enterprise system, then search the variable called

kernel.shmmax =

in /etc/sysctl.conf

Increase this value, and then save the file (as root)
Then run/sbin/sysctl Pei to activate the new configuration

SGA has been badly calibrated, causing the additional i/o or parsed hard
My SGA_TARGET was set at 1168M. I received your comments in "Enterprise Manager" - "Database Control" which gave a recommendation to increase this value to 1752M. I changed the MAX_SGA_TARGET to 1752M and also updated the SGA_TARGET 1752 m. I had to restart the instance because changing the MAX_SGA_TARGET requires it to be restarted and I could not change SGA_TARGET to 1752 until she was < = MAX_SGA_TARGET. A few hours later I went back in "Enterprise Manager" - "Database Control" to see that it was now giving a new recommendation to increase the SGA_TARGET from 1752 to 3080M.

I read that it must be true: [PGA + CMS - MEMORY_REQUIRED_BY_OS - MEMORY_REQUIRED_BY_OTHER_APPLICATIONS], TOTAL_SYSTEM_MEMORY

How will I know if the above is true in my environment? I know that my SGA is now 1752 M, my PGA is currently at 384M. How to understand MEMORY_REQUIRED_BY_OS, MEMORY_REQUIRED_BY_OTHER_APPLICATIONS and TOTAL_SYSTEM_MEMORY? This Oracle instance is the 10g version and it will run on RHEL. I was with the Linux TOP command to understand this.

On page ADDM from Enterprise Manager, this "SGA was insufficiently size" Impact find % 6.7%. Is this something I need to worry? Someone at - he saw SGA_TARGET this top in their environment?

Thanks in advance for any comments,
Wes
The second column is SGA_SIZE_FACTOR. A factor of size 1 is your current size of SGA. A size of 2 factor represents a two times larger than your current SGA SGA. A factor of size of 0.5 represents an LMS one half as great as your current size of SGA.

The fourth column is ESTD_DB_TIME_FACTOR. An ESTD_DB_TIME_FACTOR of 1 indicates that the current amount of time would be spent on physical reads. An ESTD_DB_TIME_FACTOR of 0.9355 indicates that estimates of the Oracle that you would spend 93,55% as much time on the database expects if BMG were doubled (i.e. a 6.45% saving). It's a very precise estimate, but not a very exact it is why I expanded to a range of 6 to 7%, which is even more specific, it should probably be at least translated uncertainty.

Justin

Ask the operation slow with large SGA and fast with little CMS

Hello
We have a situation where one of the insert is running slow and fast QA in PROD. Both are the same versions of database - Oracle 10.2.0.4 on HP Unix 11.31. To avoid the cause of databases running on another server, we copied from our Production database to the same server where the QA database is running and began with init.ora PROD that has 7 GB 6 GB SGA_TARGET and SGA_MAX_SIZE. For the QA database, SGA_MAX_SIZE is 700 MB and SGA_TARGET is 600 MB. Both are running on the same server, and with the same data. We have refreshed QA with data from PROD. If we start with PROD init.ora QA database, QA also behaves the same way PROD.
This problem is only with the specific insert. Here is the result of this specific statement tkprof. Can someone please interpret this for me? I am poor in SQL tuning :-( Why the statement behaves ODD with the size of the SGA PROD? Generally, we would think THAT larger SGA should give better performance.
call the query of disc elapsed to cpu count current lines
------- ------ -------- ---------- ---------- ---------- ---------- ----------
Parse 1 0.00 0.00 0 0 0 0
Run 1 56710.39 56067,75 7343 311186373 0 0
Fetch 0 0.00 0.00 0 0 0 0
------- ------ -------- ---------- ---------- ---------- ---------- ----------
Total 2 56710.39 56067,76 7343 311186373 0 0
Chess in the library during parsing cache: 1
Optimizer mode: CHOOSE
The analysis of the user id: 27 (TEST)
Rows Row Source operation
------- ---------------------------------------------------
0 CRDETAIL of SEQUENCE (cr = 0 pr = 0 pw = time 0 = 29 US)
0 REVIEWS (cr = 0 pr = 0 pw = time 0 = 21 US)
0 SORT GROUP BY (cr = 0 pr = 0 pw = time 0 = 20 US)
401 HASH RIGHT SEMI JOIN (cr = 23299915 pr 7343 pw = time = 0 = 93982966 en)
237 TABLE ACCESS BY INDEX ROWID CR_STRUCTURE_VALUES2 (cr = 96 pr = 0 pw = time 0 = 504 en)
253 CR_STRUCTURE_VALUES2_PK INDEX RANGE SCAN (cr = 4 pr = 0 pw = time 0 = 278 en)(object id 1467582)
TABLE ACCESS BY INDEX ROWID CR_COST_REPOSITORY 841 (cr = 23306003 pr 7343 pw = time = 0 = 94546465 en)
1317368058 NESTED LOOPS (cr = 79721182 pr 7343 pw = time = 0 = 18565176955 en)
VIEW 26912 (cr = pr 9874 7343 pw = time = 0 = 5269231 en)
26912 MINUS (cr = pr 9874 7343 pw = time = 0 = 5242317 en)
27462 SORT UNIQUE (cr = 9627 pr = 7329 pw = time 0 = 5040815 en)
271564 CR_STRUCTURE_VALUES2 TABLE FULL ACCESS (cr = 9627 pr = 7329 pw = time 0 = 1357961 en)
568 SORT UNIQUE (cr = 247 pr = 14 pw = time 0 = 43467 US)
TABLE ACCESS BY INDEX ROWID CR_STRUCTURE_VALUES2 2357 (cr = 247 pr = 14 pw = time 0 = US 14751)
2357 CR_STRUCTURE_VALUES2_PK INDEX RANGE SCAN (cr = pr 11 = 14 pw = time 0 = 10028)(object id 1467582) US
INDEX RANGE SCAN CRCR_MN_IX 1317341146 (cr = 79711308 pr = 0 pw = time 0 = 50420511 US)(object id 1469401)

Implementation plan of lines
------- ---------------------------------------------------
0 THE INSERT STATEMENT MODE: CHOOSE
SEQUENCE "CRDETAIL" 0 (SEQUENCE)
0 REVIEWS
0 TRI (GROUP BY)
401 HASH JOIN (RIGHT HALF)
HOW TO ACCESS THE TABLE 237: ANALYSES (BY INDEX ROWID) OF
"CR_STRUCTURE_VALUES2" (TABLE)
INDEX 253 MODE: SCANNED (SCAN INTERVAL) OF
"CR_STRUCTURE_VALUES2_PK" ((UNIQUE) INDEX)
ACCESS MODE TO THE 841 TABLE: ANALYSIS (BY INDEX ROWID) OF "CR_COST_REPOSITORY" (TABLE)
1317368058 NESTED LOOPS
VIEW 26912
26912 LESS
27462 SORT (SINGLE)
TABLE 271564 ACCESS MODE: ANALYZED (FULL) OF
"CR_STRUCTURE_VALUES2" (TABLE)
568 (SINGLE) SORT
HOW TO ACCESS THE TABLE 2357: ANALYSES (BY INDEX ROWID)
OF "CR_STRUCTURE_VALUES2" (TABLE)
INDEX 2357 MODE: SCANNED (SCAN INTERVAL) OF
"CR_STRUCTURE_VALUES2_PK" ((UNIQUE) INDEX)
MODE 1317341146 INDEX: ANALYSIS (SCAN INTERVAL) OF "CRCR_MN_IX".
(INDEX)
********************************************************************************
And here is the statement in question:
```
INSERT
INTO cr_allocations_stg
  (
    "ID",
    "COMPANY",
    "GL_ACCOUNT",
    "COST_CENTER",
    "COST_ELEMENT",
    "PROFIT_CENTER",
    "MASTER_ORDER",
    "ORDER_NUMBER",
    " FUNDING_PROJECT",
    "POSTING_ORDER",
    "POSTING_COST_CENTER",
    "ORIG_COST_ELEMENT",
    "ORIG_COST_CENTER",
    "ORIG_PROFIT_CENTER",
    "TRADING_PARTNER",
    "WORK_ORDER_NUMBER",
    "DR_CR_ID",
    "LEDGER_SIGN",
    "QUANTITY",
    "AMOUNT",
    "MONTH_NUMBER",
    "MONTH_PERIOD",
    "GL_JOURNAL_CATEGORY",
    "AMOUNT_TYPE",
    "ALLOCATION_ID",
    "TARGET_CREDIT",
    "CROSS_CHARGE_COMPANY"
  )
SELECT crdetail.nextval,
  "COMPANY",
  "GL_ACCOUNT",
  "COST_CENTER",
  '5253000',
  "PROFIT_CENTER" ,
  "MASTER_ORDER",
  "ORDER_NUMBER",
  "FUNDING_PROJECT",
  ' ',
  "POSTING_COST_CENTER",
  "ORIG_COST_ELEMENT",
  "ORIG_COST_CENTER",
  "ORIG_PROFIT_CENTER",
  " TRADING_PARTNER",
  "WORK_ORDER_NUMBER",
  CASE
    WHEN amount > 0
    THEN 1
    ELSE -1
  END,
  1,0,
  ROUND(amount * 0.0574000000, 2),
  month_number,
  0,
  '593',
  1 ,
  7,
  'TARGET',
  ' '
FROM
  (SELECT "COMPANY",
    "GL_ACCOUNT",
    "COST_CENTER",
    "PROFIT_CENTER",
    "MASTER_ORDER",
    "FUNDING_PROJECT",
    "POSTING_COST_CENTER",
    "ORDER_NUMBER",
    "ORIG_COST_ELEMENT",
    "ORIG_COST_CENTER",
    "ORIG_PROFIT_CENTER",
    "TRADING_PARTNER",
    "WORK_ORDER_NUMBER",
    month_n umber,
    0,
    SUM(amount) amount,
    SUM(quantity) quantity
  FROM CR_COST_REPOSITORY
  WHERE (amount_type    = 1 )
  AND (month_number     = 201404)
  AND ( "MASTER_ORDER" IN MASTER_ORDER
  AND EXISTS
    (SELECT 1
    FROM
      (SELECT SUBSTR(ELEMENT_VALUE, 1, DECODE(INSTR(ELEMENT_VALUE, ':'), 0, L ENGTH(ELEMENT_VALUE) + 1, INSTR(ELEMENT_VALUE, ':')) - 1) AS ELEMENT
      FROM CR_STRUCTURE_VALUES2
      WHERE STRUCTURE_ID       = 2
      AND DETAIL_BUDGET        = 1
      AND STATUS               = 1
      AND UPPER(PARENT_VALUE) IN ('ELECTRIC ALL OTHER','ELECTRIC COR')
      MINUS
      SELECT SUBSTR(ELEMENT_VALUE, 1, DECODE(INSTR(ELEMENT_VALUE, ':'), 0, LENGTH(ELEMENT_VALUE) + 1, INSTR(ELEMENT_VALUE, ':')) - 1) AS ELEME NT
      FROM CR_STRUCTURE_VALUES2
      WHERE STRUCTURE_ID      = 9
      AND DETAIL_BUDGET       = 1
      AND STATUS              = 1
      AND UPPER(PARENT_VALUE) = 'A&G OH ORDER EXCLUSION'
      ) Z
    WHERE Z.ELEMENT = MASTER_ORDER
    )
  AND "GL_ACCOUNT"   <> '91081001'
  AND "COST_ELEMENT" IN COST_ELEMENT
  AND EXISTS
    (SELECT 1
    FROM CR_ST RUCTURE_VALUES2 A
    WHERE A.STRUCTURE_ID = 5
    AND A.DETAIL_BUDGET  =1
    AND A.STATUS         = 1
    AND COST_ELEMENT     = A.ELEMENT_VALUE
    )
  AND "GL_ACCOUNT" NOT IN ('5100000','5325000','5327000')
  AND "SOURCE_ID"      <> '7' )
  GROUP BY "COMPANY",
    "GL_ACCOUNT",
    "COST_CENTER",
    "PROFIT_CENTER",
    "MASTER_ORDER",
    "FUNDING_PROJECT",
    "POSTING_COST_CENTER",
    "ORDER_NUMBER",
    "ORIG_COST_ELEMENT",
    "ORIG_COST_CENTER",
    "ORIG_PROFIT_CENTER",
    "TRADING_PARTNER",
    "WORK_ORDER_NUMBER",
    month_number
  )
```
Enjoy your first answer on this.
Thank you and best regards,
Murali

Option 1:

You run with two different ORACLE_HOMEs - that is potentially two different copies of the Oracle executable to the patch different sets?

Option 2:

A change of size of SGA is unlikely to directly affect the execution plan, but it also has change the size of the PGA TOUR at the same time? A change in the pga_aggregate_target could affect the choice of the mechanism of the join optimizer. (But not to change an in a join; but hash semi-join to nested loop is possible).

Possibility 3:

The db_file_multiblock_read_count leaves then the size that oracle defines by default depends on the db_cache_size divided by process; so, if you have reduced the sga_target_size you (implicitly or explicitly, no doubt) reduced the db_cache_size, and if you don't reduce the process in the same way, then the default db_file_multiblock_read_count been reduced. What you did on your system stats, this could change the cost of (for example) full tablescans, which could lead to a change in execution plan.

It would be useful to see the results of a call to explain the plan / dbms_xplan.display in both cases so that we can see the variation in estimates of Oracle.

Concerning

Jonathan Lewis

standby db memory size
Hello

It's a simple question: when I have an instance ensures Data Guard, must have the same size of SGA/PGA as main? Or can it be reduced and if so, are there limits (according to the primary workload ofc). What would be an appropriate relationship between primary: secondary pga/sga?

Thank you.
Mika wrote:
Hello

It's a simple question: when I have an instance ensures Data Guard, must have the same size of SGA/PGA as main? Or can it be reduced and if so, are there limits (according to the primary workload ofc). What would be an appropriate relationship between primary: secondary pga/sga?

Thank you.

No didn't need configuration of memory set to set as primary. But you must set the SGA/PGA correctly, because when you apply archive logs you can get errors (if you don't not gen so that all errors are ok.)

Automatic shared memory management
Hi, the log buffer is not the interior components of size automatically sga?
Please refer to the documentation
http://download.Oracle.com/docs/CD/B19306_01/server.102/b14220/memory.htm#CNCPT810

And also, before ask you a question, do a simple search in google or tahiti.oracle.com :)

- - - - - - - - - - - - - - - - - - - - -
Kamran Agayev a. (10g OCP)
http://kamranagayev.WordPress.com
[Step by step installation Oracle Linux and automate the installation by using Shell Script | http://kamranagayev.wordpress.com/2009/05/01/step-by-step-installing-oracle-database-10g-release-2-on-linux-centos-and-automate-the-installation-using-linux-shell-script/]

Published by: Kamran Agayev, a., 16 July 2009 15:37

Questioning ADDM recomentation memory

Hi all
I was going by recommendation ADDM and came across this:
Results: The SGA has been badly calibrated, causing additional i/o or hard analysis
Action: Increase the size of the SGA by setting the parameter "30720M.
Size of SGA is currently 20395M
I checked use SGA and PGA for the last month:
-Gets number of hits when the memory usage is less than 5% of memory available max LMS
Select SNAP_ID, INSTANCE_NUMBER, USED_MB
FROM (select SNAP_ID, INSTANCE_NUMBER, sum (round(bytes/1024/1024,0)) USED_MB
of DBA_HIST_SGASTAT
WHERE name! = 'free mΘmoire '.
Group of SNAP_ID, INSTANCE_NUMBER)
WHERE USED_MB > (by selecting round ((sum (VALUE) / 1024/1024/100) * 95, 0) in v$ sga)
/
no selected line
-Gives me max using SGA for the last month
Select INSTANCE_NUMBER, max (USED_MB)
FROM (select SNAP_ID, INSTANCE_NUMBER, sum (round(bytes/1024/1024,0)) USED_MB
of DBA_HIST_SGASTAT
WHERE name! = 'free mΘmoire '.
Group of SNAP_ID, INSTANCE_NUMBER)
INSTANCE_NUMBER group
/
INSTANCE_NUMBER MAX (USED_MB)
--------------- ------------
2 19002
1 18783
3 18985
If BMG never went more 18985 M instance within the last month
Then I started thinking about the PGA
Size of the PGA is currently at 4 GB
Same set of queries (but with PGA):
-Gets number of hits when the memory usage is less than 5% of memory available max PGA
Select SNAP_ID, INSTANCE_NUMBER, USED_MB
FROM (select SNAP_ID, INSTANCE_NUMBER, sum (round(value/1024/1024,0)) USED_MB
of DBA_HIST_PGASTAT
where name = 'total inuse PGA.
Group of SNAP_ID, INSTANCE_NUMBER)
WHERE USED_MB > (select round ((sum (VALUE) / 1024/1024/100) * 95, 0) from v$ pgastat, whose name = "aggregate target PGA parameter")
/
Number of rows returned here...
-Gives me max use PGA for the last month
Select INSTANCE_NUMBER, max (USED_MB)
FROM (select SNAP_ID, INSTANCE_NUMBER, sum (round(value/1024/1024,0)) USED_MB
of DBA_HIST_PGASTAT
where name = 'total inuse PGA.
Group of SNAP_ID, INSTANCE_NUMBER)
INSTANCE_NUMBER group
/
INSTANCE_NUMBER MAX (USED_MB)
--------------- ------------
3-1553
2 1649
1 7877
Instance etc. we PGA TOTAL running went in close to 8 GB
Should I be not increasing the PGA rather than the LMS?
If my select statements are incorrect please let me know
Concerning
G.

Understood why Oracle recommends SGA increase

Parsed hard on nodes 1 and 2 were ok, 3.8% and 0.06% respectively

But in node 2, hard analysis is at 12.1%

As a result, the oracle of reason recommended this.

I still don't understand why the use of the LMS was not maxed out, however.
I would have thought that:

Select SNAP_ID, INSTANCE_NUMBER, USED_MB

FROM (select SNAP_ID, INSTANCE_NUMBER, sum (round(bytes/1024/1024,0)) USED_MB

of DBA_HIST_SGASTAT

WHERE name! = 'free mΘmoire '.

Group of SNAP_ID, INSTANCE_NUMBER)

WHERE USED_MB > (by selecting round ((sum (VALUE) / 1024/1024/100) * 95, 0) in v$ sga)

/

Would have returned at least one line.
Maybe I need to narrow it down to 90% instead of 95%

In any case I'll add memory on all nodes

For all those who are interested:

SQL > INST_ID select, name, value

2 gv$ sysstat

3 WHERE name = 'parse count (total)'

4 or name = 'parse count (hard).

5.

INST_ID NAME VALUE

---------- ---------------------------------------------------------------- ----------

2 analysis count 73538108 (total)

2. analyze the County 36760147 (hard)

3 analysis count 152131576 (total)

3 analysis County 114045 (hard)

1 analysis count 860962865 (total)

1 analysis County 11106572 (hard)

6 selected lines.

SQL > exec DBMS_LOCK. SLEEP (360);

PL/SQL procedure successfully completed.

SQL > INST_ID select, name, value

2 gv$ sysstat

3 WHERE name = 'parse count (total)'

4 or name = 'parse count (hard).

5.

INST_ID NAME VALUE

---------- ---------------------------------------------------------------- ----------

3 analysis count 152184466 (total)

3 analysis County 114082 (hard)

2 analysis count 73538835 (total)

2. analyze the County 36760235 (hard)

1 analysis count 860980050 (total)

1 analysis County 11107238 (hard)

6 selected lines.

Definition of size of SGA in 10g

Similar Questions

Maybe you are looking for