PGA and SORT_AREA_SIZE

Similar Questions

• Metric of the influence of the size of the PGA and UGA County cursors

  Hello!
  
  If I have the right to understand,
  the child cursors County growthing indirectly affects the size of the PGA and UGA,
  If that's true, how to receive absolute metric (* GA, no Shared Pool)?
  
  
  Thank you
  Paul
  
  Published by: Paul on 06.02.2012 21:28

  Hi Paul,.

  To be more precise, I refer to the implicit cursors that are stored in the library cache. Please do not confuse with the cursors that are defined explicitly in pl/sql.

  For each identical sql statement, oracle will create the shared cursor (library cache), if this is the first time that the statement is then executed as said previously it will create the cursor in LB, so is the second time, the same SQL statement is executed then oracle will share already created the cursor to the second statement.
  Under normal circumstances for each identical instruction, there are 1 parent and 1 child of the cursor.

  Consider the example next where the sql statement appears to be the same but do not share already created cursor and therefore it must create a new.

  Session executed following queries:

  Select the name of the employee; -> It is that first time sql is executed, then oracle will open a new cursor. In a specific cursor, you have the code sql and other execution information.

  Session B executed after the request:

  Select the name of the employee; -> Here statement SQL makes sense even with above but this will create a new cursor, so already open for session A will not be shared. Note here that the SELECT is uppercase, then oracle think it's the different statement and thus create a new cursor.

  It's an example where difference in uppercase letter oracle thinks it's a new statement. But there are other schenarios where the same situation occurs.
  Another reason for cursor not shared are in V$ SQL_SHARED_CURSOR.

  Although oracle database running in the situation where the sliders are not shared, then this can lead to performance impact hugh like lock in shared long, shared pool leak memory pool, library catch latch etc.. More important, which will require another cursor important to eliminate memory to place a new

  I hope you already know, what is analysis difficult. If this increase the situation of hard analysis, which may further affect the performance.

  Let me know in case of doubts

  Concerning

  Published by: 909592 on 7 February 2012 13:46

  Published by: 909592 on 7 February 2012 13:49

• PGA and SGA sizing on linux

  Oracle 11.0.1.7 on linux 64-bit:
  
  Just a basic question about the PGA and SGA. Our SGA_MAX is 9 GB and PGA_AGGREGATE_TARGET is 4800M, so I expect the release of 'ipcs-a' to 9 GB + 4800 M. but it's actually only 9GB. When I query V$ PGASTAT it shows that he attributed to 4800M. So this means that the shared memory is not used for this purpose? It's all in the RAM? I'm trying to calculate what my SGA and PGA should be about 16 GB of RAM where Oracle is the dominant process only. PGA_ADVICE said that 3600 M is optimal. So is it OK to use 11 + 3600 M and leave just 1 GB for OS?
  
  I am also looking to allow the HugePages I see many pages is paged out.

  The SGA_ADVICE column is based on real activity, not coming out of the operating system to determine the amount of RAM is available and protective in a reasonable allocation of SGA and PGA. If your database was not yet under a lot of load, SGA_ADVICE will tell you that you could radically reduce the size of the SGA without substantial effect on performance. Of course, when you add load / add a volume of data, that advice will change usually fairly quickly.

  Justin

• SGA, PGA and other areas of memory

  Hi guys,.
  
  (I use 9i, with dedicated server mode, for a DSS system)
  
  I just want to know if my interpretation is correct.
  
  (1) when an instance starts, Oracle sets aside the amount of RAM specified by SGA_MAX_SIZE.
  (2) given that the PGA is located outside the LMS, Oracle also cancels PGA_AGGREGATE_TARGET.
  (3) so that the required total RAM is SGA_MAX_SIZE + PGA_AGGREGATE_TARGET
  (4) the SGA has a variable size. Because stuff like SHARED_POOL_SIZE and JAVA_POOL_SIZE may have a value specified, their size cannot grow. Therefore, the variable size is used to account for the other structures of memory as sql_area I can see on the page of Tom Kyte on SGA (links not working do not, so the address is http://asktom.oracle.com/pls/asktom/f?p=100:11:0:::P11_QUESTION_ID:365088445659))
  (5) If a database is about 85% of the SGA size is, would you say that is a size very bad CMS? (I know you guys hate rules of thumb, but any input on this would be always appreciated).
  
  Thank you

  It's confused me, I thought that each user is able to use up to the specified amount... If a user can go beyond PGA_AGGREGATE_TARGET, what's the point of this setting? Of course, Oracle must have an upper limit for the PGA, otherwise it could eventually consume all memory.

  First confusion PGA_AGGREGATE_TARGET is not by user, is limited for all processes associated with the instance server. Always book enough memory on the OS to ensure that you can honor the value.
  http://68.142.116.68/docs/CD/B19306_01/server.102/b14237/initparams157.htm#REFRN10165

  There are cases where Oracle on will provide PGA if PGA_AGGREGATE_TARGET set too low.
  For example.

  SELECT * FROM V$PGASTAT;The output of this query might look like the following:
  
  NAME                                                          VALUE UNIT-------------------------------------------------------- ---------- ------------aggregate PGA target parameter                             41156608 bytesaggregate PGA auto target                                  21823488 bytesglobal memory bound                                         2057216 bytestotal PGA inuse                                            16899072 bytestotal PGA allocated                                        35014656 bytesmaximum PGA allocated                                     136795136 bytestotal freeable PGA memory                                    524288 bytesPGA memory freed back to OS                              1713242112 bytestotal PGA used for auto workareas                                 0 bytesmaximum PGA used for auto workareas                         2383872 bytestotal PGA used for manual workareas                               0 bytesmaximum PGA used for manual workareas                       8470528 bytesover allocation count                                           291bytes processed                                          2124600320 bytesextra bytes read/written                                   39949312 bytescache hit percentage                                          98.15 percent
  

  the count of distribution: this statistic is cumulative since instance startup. Too PGA Award in memory can occur if the value of PGA_AGGREGATE_TARGET is too small to accommodate the other PGA component in the previous equation, more the minimum memory required to run the load of work area. In this case, Oracle cannot honor PGA_AGGREGATE_TARGET initialization parameter and additional PGA memory should be allocated. In the case of over-allotment, you must increase the value of PGA_AGGREGATE_TARGET using the information provided by the view of advice V$ PGA_TARGET_ADVICE.

  http://68.142.116.68/docs/CD/B19306_01/server.102/b14211/memory.htm#i49320

  I'm still very confused about SHARED_POOL_SIZE. If I put SHARED_POOL_SIZE to 100M, and according to you, guys, that can grow to 100 M, so what's the point of having this setting in the first place?

  Thank you

  No, shared pool does not in itself, unless you use alter system to change. It's the dynamic sense, you don't need to restart the example to change. In the earlier version, you must restart the instance to resize the shared pool and the buffer cache.

  Oracle makes this possible by introducing setting SGA_MAX_SIZE in 9i.
  Before 9i, SGA is essentially BUFFER_CACHE + SHARED POOL + a few other components. Oracle allocate memory from the OS at startup function db_block_size and db_block_buffers shared_pool_size etc. So change the any of these settings for example, must restart.
  With SGA_MAX_SIZE, Oracle will allocate always SGA_MAX_SIZE initially, for example, 500M, you can define db cache and share setting as small as 200 M + 200 M pool, the 100M unused will be reported as free CMS, that allow to increase the shared pool and buffer cache later and also to provide a hold space released place if these setting reduced.

• question about the use of the PGA and TEMP

  Hello
  We had a situation whereby OEM reported that because the PGA was too small for a period of time that e/s additional on the TEMP was created. So does that mean that the TEMP has acted as a pseudo PGA.
  
  concerning

  Take a look at these threads
  Re: Problem with memory PGA - Oracle 10.2.0.4 on windows 2003
  Re: Details on using in the memory of the PGA

  
  
  Oracle database FAQ
  
  

• Maximum session reached. Y at - it report any with PGA and SGA?

  Friends,
  
  OS: RHEL AS 3
  DB: 9iR2
  
  Currently...
  Process = 150
  sessions = 170
  
  now, I want to be
  Process = 250
  sessions = 280
  
  my current sysctl.conf is...
  

  # Oracle 9i kernel parameters
  #
  kernel.sem = 250 32000 100 150
  kernel.shmmax = 4294967295
  kernel.shmmni = 4096
  kernel.shmseg = 4096
  kernel.shmall = 3279547
  fs.file-max = 327679
  kernel.msgmni = 2878
  kernel.msgmnb = 65536
  net.ipv4.ip_local_port_range = 1024 65536
  #
  # End of parameters

  SQL> select * from v$sga;
  
  NAME                      VALUE
  -------------------- ----------
  Fixed Size               452164
  Variable Size         268435456
  Database Buffers      369098752
  Redo Buffers             667648

  and our server ram is 2 GB.
  
  is that the values above are sufficient to improve processes and sessions for 250 and 280?
  What are the limits to improve processes and sessions?
  I think also, long back I asked the same question. but here, I searched is not here. After the upgrade of this forum this thread has been deleted.
  
  Thank you

  VPS wrote:
  Friends,

  Currently...
  Process = 150
  sessions = 170

  now, I want to be
  Process = 250
  sessions = 280

  The process parameter controls a number of allowances in the SGA heap table - (which is the array of sessions, by default, processes * 1.1 + 5). An increase in the process automatically increases the variable size in 9i.

  But here are some numbers to a 32-bit process = 150 Windows startup
  125829120 variable size bytes

  then, with the process = 250
  Variable 134217728 bytes size

  As you can see, the additional allowance is only about 8.5 MB - and this does not to cause any problems with your settings in memory.

  There is an informal indication of 10g, you should have a db_cache_size which is at least 1 MB per process - and you are already experiencing this figure - even after the increase.

  In a reasonable manner for the pga_aggregate_target for systems OLTP is put somewhere between 1 MB and 4 MB per process - depending on the nature and the amount of SQL statements used. If you do have enough memory to deal with it (as long as you don't get too close to the end of 4 MB of the scale - v$ v sql_workarea_histogramand $ pgastat to see how your current pga_aggregate_target is to check).

  Process more means more open cursors, however, you may have a relatively small increase in the shared_pool_size if your application is good for sharing SQL - but a larger increase if your code is not very common.

  You can also look at an article and a presentation by Christo Kutrowski of the Pythian Group which deals related to Oracle o/s overhead, its SGA and the number of process - and how to minimize.
  See: http://www.pythian.com/blogs/741/pythian-goodies-free-memory-swap-oracle-and-everything

  Concerning
  Jonathan Lewis
  http://jonathanlewis.WordPress.com
  http://www.jlcomp.demon.co.UK

  "The greatest enemy of knowledge is not ignorance, it is the illusion of knowledge." Stephen Hawking.

• What is sga and pga configuration set up

  Hi Expert,
  
  Please help me with my question
  
  What is the exact configuration sga, pga and other setting
  
  Win2003 operating system 32-bit and oracle 8i
  
  64 GB of RAM
  database size 70 GB
  processor 2.13 ghz
  
  BMG =
  PGA =
  db_block_buffers =
  shared_pool_size =
  LARGE_POOL_SIZE =
  log_buffer =
  DB_BLOCK_SIZE =
  sort_area_size =
  sort_area_retained_size =

  user12255952 wrote:
  my hit ratio is low 76
  So what can I do

  What ratio of hit?

• Question of PGA

  Hello

  Oracle 11.2.0.3.0

  I have memory_target = 15G

  That is why in the database, pga_aggregate_target = 0

  But I can see from v$ pgastat, aggregation PGA target = 5637144576 parameter = Mo 5673 .

  Statspack shows

  Statistical memory Begin End

  ~~~~~~~~~~~~~~~~~                ------------ ------------

  Host Mem (MB): 31,440.3 31 440,3

  Use of LMS (MB): 15,291.4 15 291,4

  PGA use (MB):                          1,722.1      1,698.8

  Host % Mem used for SGA + PGA: 54.0 54.1

  PGA Aggr target histogram DB/Inst: PIVR01/PIVR01 snaps: 8610-8612

  -> Optimal performances are purely in memory operations

  High low

  Execs Total optimum optimum optimum Execs 1 pass Execs M-Pass Execs

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

  2K      4K                     5,036          5,036            0              0

  64K    128K                       14               14            0              0

  128K    256K                      78              78             0              0

  512 K 1024 K 174 174 0 0

  1M      2M                      164            164            0              0

  4M      8M                           4               4             0              0

  256M 512M                      2               1             1              0

       1G      2G                             2               0             2              0

  
  

  Memory PGA Advisory DB/Inst: PIVR01/PIVR01 end Snap: 8612

  --> When using automatic memory management, choose at least a value of pga_aggregate_target

  where Estd PGA Overalloc Count is 0

  Estd estd Extra

  AGGR PGA W/A MB time Estd Estd PGA PGA

  Target size W/A MB read/written to deal with Cache Overalloc

  Factr EST (MB) treated disc bytes (s) % number of access

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

  672 0.1 427 978 214.7 522 063 45.0 3 481

  1 344 0.3 427 978 514 547 213,0 45.0 3 320

  2 688 0.5 427 978 90 681 117.2 83,0 0 * advised a minimum size to fit all executions in memory        

  4 032 0.8 427 978 84 797 115.9 83,0 0

  5 376 1.0 427 978 115,3 82 105 84.0 0 * size current target which we fail to

  6 451 1.2 427 978 114,0 76 674 85.0 0

  7 526 1.4 427 978 114,0 76 674 85.0 0

  8 602 1.6 427 978 114,0 76 674 85.0 0

  
  

  1. why PGA memory grows no more than 2 GB to cope with this and be able to run the SQL all in memory, even if the target is set at 5673 MB?

  2. do I need to manually set pga_aggregate_target = 2688 M with memory target = 15 G?

  
  

  Thanks for your time
  

  1938071 wrote:

  I got your point, but the bottom implies that there is 2 SQL executions which do not completely held in memory of the PGA and must be partially exchanged on temporary slow disk. The size of the book that does this is between 1-2 GB.

  So if I 5673 MB allocated total setting target PGA in v$ pgastat, why is he not use it... This part confuses me

  Target histogram Aggr PGA

  -> Best executions are purely in memory operations

  High low

  Execs Total optimum optimum optimum Execs 1 pass Execs M-Pass Execs

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

  2K      4K                     5,036          5,036            0              0

  64K    128K                       14               14            0              0

  128K    256K                      78              78             0              0

  512 K 1024 K 174 174 0 0

  1M      2M                      164            164            0              0

  4M      8M                           4               4             0              0

  256M 512M                      2               1             1              0

       1G      2G                             2               0             2              0

  The pga_aggregate_target did NOT previously assigned, it is a (soft) limit on the total memory that can be allocated dynamically. Given that you have defined the memory_target to 15 GB, Oracle can only grow the total allocation of the PGA to 5G by reducing the rest of its endowment (in large part the shared pool and the buffer cache) to 10 GB.

  It is a necessary part of the algorithm of automatic memory management that it would decide if the time saved by the growth of the component of the PGA is greater or less than the time lost by the decline of the shared pool and the cache buffers.  You could imagine that doing a few GB of bed and writes temporary files for one or two is probably much faster that 2 GB buffer cache dumping and ending with a large number of random to block reads as a side effect.

  -If you have a CPU bottleneck then hit a very large size on disk may be the best use of resources: large sorts in memory can we a lot of CPU.

  Concerning

  Jonathan Lewis

• Merging databases and memory settings

  I have two separate long-term database server (11r2 on RHEL 5). We seek to combine them into a single server in order to reduce the costs and delays of the admin. On both servers, I can run...

  select * from v$sga_target_advice order by sga_size;
  

  .. .and get plenty of data on each box. I can check the current sizes for parameters static EAMA does not dynamically as well.

  Question: if I'm their fusion, the initial memory parameters (sga_target, etc.) must simply be the sizes from two servers added?

  Aka, if SGA_TARGET is 4 GB on db1 and 5 GB on db2, the combined server is 9 GB?

  Thank you

  You should study hugepages.

  Sybrand is correct.  In a dynamic system with many variables, the merger of the two systems is not additive.  You go through an iterative process of adjustment.  Even the advisers are only a starting point, they tend to want to, just to keep memory more.  Find the paper to think clearly on Cary Millsap's performance and consider there are probably things that are fine when separate them, but you cut above the knees when combined.  Unless, of course, the two systems are currently passing too provisioned.

  You should be able to anticipate some things by taking the CWA (if authorized, statspack if not) reports to see what is normal.  Keep just a few looks at the pga and sga statistics.  Redo the calibration is probably something that is additive.

• Part of the SORT_AREA_RETAINED_SIZE of the UGA in shared server?

  Hello
  
  I'm confused.
  
  Here are the notes I made during the expert reading by Tom Kyte Oracle a few years ago.
  
  SORT_AREA_SIZE - the amount of memory to use for sorting. Will be in the PGA
  SORT_AREA_RETAINED_SIZE - how much memory is used to store the sorted data. Will be in the UGA
  
  I now see the notes that I did as I was not working with Oracle for quite awhile, and I don't know what is happening in shared server mode.
  
  SORT_AREA_RETAINED_SIZE would still be part of the UGA? Probably not, as S_A_R_S is a subset of S_A_Z, Yes?
  
  Thank you.

  >
  SORT_AREA_RETAINED_SIZE would still be part of the UGA? Probably not, as S_A_R_S is a subset of S_A_Z, Yes?
  >
  Since the doc should answer your question
  >
  SORT_AREA_RETAINED_SIZE specifies (in bytes) the maximum amount of memory of the global area (UGA) user retained after that complete a sort of race. Selected size the size of the buffer control learn that Oracle uses to maintain some of the sort in memory. This memory is released to the UGA, not for the operating system, after the last line is read in the sorting space.

  Oracle can allocate sort on several areas of this size for each request. Usually, only one or two kinds occur at the same time, even for complex queries. In some cases, however, additional concurrent sorts are necessary, and each kind has its own memory space. If shared server is used, allowance is BMG until the SORT_AREA_RETAINED_SIZE value is reached. The difference between SORT_AREA_RETAINED_SIZE and SORT_AREA_SIZE is allocated to the PGA.
  >
  SARS is preserved UGA (first line above) and the difference between SARS and SAS PGA (last line above).

  Oracle recommend to use one of these settings more
  >
  Note:

  Oracle does not recommend the use of the SORT_AREA_RETAINED_SIZE parameter, unless the instance is configured with the option of shared server. Oracle recommends enable auto sizing of SQL working areas by setting PGA_AGGREGATE_TARGET instead. SORT_AREA_RETAINED_SIZE is retained for backward compatibility.
  >
  SORT_AREA_SIZE
  http://docs.Oracle.com/CD/B19306_01/server.102/b14237/initparams203.htm

  SORT_AREA_RETAINED_SIZE
  http://docs.Oracle.com/CD/B28359_01/server.111/b28320/initparams231.htm

• Full Temp tablespace

  Hello
  
  The tablespace temp on our database suddenly began to fill.
  
  It has 4 GB of ram on your server, about 1.3 G is used by the operating system,
  , 1.7 bg the sga and 700 MB by PGA target parameter.
  
  We use the DSO, then the sort_area_size parameter does not come into play.
  
  All ratios are fine, but once the temp tablespace is almost full, things grind to stop.
  
  Someone at - it advice on the best way forward.
  
  I put statspack to run at the time.
  
  I appreciate ram is low, but I am confused as to why he suddenly began to occur.
  
  Thanks for reading.

  PGA_TARGET is a target.

  fixed lines 200
  Select * from v_$ pgastat;

  and look at the maximum used.

  There are also counselors, that you can watch in the views or dbconsole (if permitted), you can watch advice of the pga and pga memory utilization details. If you do a lot of executions multipass, which would slow and use a lot of temp. Temp tends to stay fully, docs of MOS and comments here that explain it.

  Adds the numbers you have validated, looks that you run right up to the border in normal times, a little more use of the user and you jump off the cliff. You may need to reduce your sga. How many users are using a shared server, what happens to change things?

• Great UGA in a dedicated DB connections

  Hi all
  
  I got the total PGA and UGA memory used by the sessions of the request below. Connections used dedicated connections.
  
  My question is: why the UGA is that big (near total PGA)? Is it normal for a database that has no connection to the shared server. What is the reason for this?
  
  Oracle 11g R1
  Windows 2003 SP2
  

  SQL> SELECT n.name, round(SUM(value)/1024/1024,2) MB
    2  FROM V$SESSTAT s, V$STATNAME n
    3  WHERE s.statistic# = n.statistic#
    4  AND n.name in ('session pga memory')
    5  group by n.name
    6  union
    7  SELECT n.name, round(SUM(value)/1024/1024,2) MB
    8  FROM V$SESSTAT s, V$STATNAME n
    9  WHERE s.statistic# = n.statistic#
   10  AND n.name in ('session uga memory')
   11  group by n.name
   12  ;
  
  NAME                                                         MB
  ---------------------------------------------------- ----------
  session pga memory                                       816.82
  session uga memory                                       782.36

  It is normal that the UGA to make up most of the PGA TOUR. The UGA includes the sorting of the user box.

  What type of memory management you used? What are the values for the parameters of database: sort_area_size and sort_area_retained_size?

  HTH - Mark D Powell.

• Issue update Compaq 6510 processor b

  Hello
  
  I'm about to upgrade the laptop processor for my Compaq 6510 b-computer that currently has the Intel Celeron 540 (1.86 GHz) processor. I studied the latest international document quickspecs (p. 4), which says that the laptop must be compatible with the Intel Core 2 Duo T7700, which I would be interested in upgrading to.
  
  However, I noticed that there are 4 different "versions" of this processor on the market of the product descripition Intel. These "versions" are called: SLA43, SLAF7, SLA3M and SLADL. Now I'm about to buy the tray (no not in box) SLAF7-model, but would like confirmation that this model works with the Compaq 6510 b-computer laptop?

  I did some additional research on the 4 models and thought I'd do a short summary on them even, because someone might find it useful. From my understanding and looking at the "'Ordering / sSpecs / Steppings"-page on the left side of theIntel product description page, both latter models (SLA3M, SLADL) are models of Ball Grid Array (BGA), which means they are soldered on the motherboard and so I can't use them for an upgrade. " However, the SLA43 and SLAF7 models are the two tables in grid Pin (PGA) and so I should be able to upgrade to one of these. The only difference between the SLA43 a SLAF7 of models is that they have different steps: "E1" and "G0". However, it seems that in practice this difference in step by step is an important place. Please correct me if I'm wrong.
  
  Thank you in advance for your help!

  The best way to get more information is by checking the specs on the numbers for INTEL processors optionally simplify research candidate upgrade.

  Intel Core Duo T7700 2.4 GHz with 4 MB of cache L2 446894-001

  It can be found at a reasonable price. I found one was offered at $60.00 USD.

  http://www.serversupply.com/processors/Intel%20Core%202%20Duo/2.4GHz%20-%20800MHZ%20FSB/HP/446894-001.htm

  

• Recommendations of the RAM to 7000 concurrent connections

  Hi all

  I need your kind recommendations.

  We have installed oracle database enterprise edition 12 c on windows server 2012. the two 64-bit.

  We have 14000 students. We want to provide to 7000 concurrent user connections.

  How much RAM we forced to 7000 concurrent connections?

  What CPU is necessary?

  Concerning

  You have 2 possibilities to process client connections in the database - dedicated server and shared server (or a combination of both).

  Shared server has a smaller footprint resource wise. In dedicated server, each client connection is served by a dedicated server process. With a shared server, something like a 100 customer connections can be repaired by a pool of 10 shared server process. Shared server requires however a larger SGA as the UGA for a client connection must reside in the global shared memory, and not local in the process of shared server maintenance in this regard. Also, shared server is only really suitable for OLTP and not the customers of type OLAP.

  Shared server can drastically reduce the footprint of server for a large client connection resource base - if configured and used correctly.

  As for Windows. Oracle uses Windows threads. A process unique oracle.exe stops, and all the system processes (process monitor, writer of the db, etc.) and all (dedicated and shared server process) user processes running such as discussions in this process of exe. A thread has a minimum size imposed by the kernel - and a default size (often larger than the minimum) is defined by the process that created the thread.

  The last time I ran Oracle on Windows has been with Oracle 7.1 - so I'm not at all recent with Oracle running on the NT kernel. But I remember correctly, the Oracle default threads thread size is 2MB. Determine the current size and plug that number into the following discussion:

  So for a dedicated server, you'll need 7000 x 2 MB of memory for threads in dedicated server - which will develop according to the UGA and needs PGA of the session of customer served by this thread.

  If is considered shared server, you can try a pool shared of a 1000 servers, with a 100 dispatchers - (1000 + 100) x 2 MB of memory. In this case, the server shared memory can develop due to the needs of PGA. The UGA lies in the SGA however (so a larger SGA is required with dedicated server where the UGA is local and resides in the PGA).

  The average size of the PGA and UGA can be complex - because it depends entirely on the running of this process (shared/dedicated) server code. Lots of bulk code treatment? You can expect PGA and ramp up a lot as a binder in bulk and extraction in block needs memory to the process.

  Whatever it is, you need a 64-bit platform to a 64-bit operating system.

  Issue. Why Windows as a server database? Oracle Linux as the operating system is in my view, greater flexibility of choice ito, scalability and costs.

• cursors of misunderstandings

  I have a bit of a mess in my head when I read on the sliders. It seems to me that the term has often loosely to describe things a little different. Sometimes, I read on the pointer cursor. Sometimes, as an area of memory. Sometimes, it is created in the PGA and sometimes SGA suggesting that is actually is pointing on execution plan... I understand that more advanced users use the military when talking about things related slider, and they understand what they are talking about, but for a beginner, these shortcuts are really confusing.

  Could you give me some advice on recommended books/articles/chapters describing cursors that are precise, step by step and well defined terms to distinguish the different things that can go under the term common "cursor"?

  Thank you!

  Yes, you get the idea base.

  It can be however "special areas for the data in row", created by a cursor - usually for the purpose of sorting or grouping. However, these aren't use to cache the rows as a result set memory of all kinds. What we called areas of activity in the old days of mainframe language. The cursor will allocate temporary space for these areas (one of the reasons for the existence of the TEMP tablespace) - and this temp space can contain data of actual line required for processing.

  However, the perception that a cursor is defined in a result in memory is problematic - as this concept will not scale. How many sessions will be able to open a cursor line 10,000 (with each line containing a CLOB from an average of 1 MB), before the server memory is exhausted, and core is swapping virtual pages like crazy, without real CPU time left for anything else running?