8 Memory Architecture

The SGA_TARGET Initialization Parameter

The SGA_TARGET initialization parameter reflects the total size of the SGA and includes memory for the following components:

• Fixed SGA and other internal allocations needed by the Oracle Database instance

• The log buffer

• The shared pool

• The Java pool

• The buffer cache

• The keep and recycle buffer caches (if specified)

• Nonstandard block size buffer caches (if specified)

• The Streams pool

It is significant that SGA_TARGET includes the entire memory for the SGA, in contrast to earlier releases in which memory for the internal and fixed SGA was added to the sum of the configured SGA memory parameters. Thus, SGA_TARGET gives you precise control over the size of the shared memory region allocated by the database. If SGA_TARGET is set to a value greater than SGA_MAX_SIZE at startup, then the latter is bumped up to accommodate SGA_TARGET.

Automatically Managed SGA Components

When you set a value for SGA_TARGET, Oracle Database 10g automatically sizes the most commonly configured components, including:

• The shared pool (for SQL and PL/SQL execution)

• The Java pool (for Java execution state)

• The large pool (for large allocations such as RMAN backup buffers)

• The buffer cache

• The Streams pool

You need not set the size of any of these components explicitly. By default the parameters for these components will appear to have values of zero. Whenever a component needs memory, it can request that it be transferred from another component by way of the internal automatic tuning mechanism. This transfer of memory occurs transparently, without user intervention.

The performance of each of these automatically sized components is monitored by the Oracle Database instance. The instance uses internal views and statistics to determine how to distribute memory optimally among the components. As the workload changes, memory is redistributed to ensure optimal performance. To calculate the optimal distribution of memory, the database uses an algorithm that takes into consideration both long-term and short-term trends.

Manually Managed SGA Components

There are a few SGA components whose sizes are not automatically adjusted. The administrator needs to specify the sizes of these components explicitly, if needed by the application. Such components are:

• Keep/Recycle buffer caches (controlled by DB_KEEP_CACHE_SIZE and DB_RECYCLE_CACHE_SIZE)

• Additional buffer caches for non-standard block sizes (controlled by DB_nK_CACHE_SIZE, n = {2, 4, 8, 16, 32})

The sizes of these components is determined by the administrator-defined value of their corresponding parameters. These values can, of course, be changed any time either using Enterprise Manager or from the command line with an ALTER SYSTEM statement.

The memory consumed by manually sized components reduces the amount of memory available for automatic adjustment. For example, in the following configuration:

SGA_TARGET = 256M
DB_8K_CACHE_SIZE = 32M

The instance has only 224 MB (256 - 32) remaining to be distributed among the automatically sized components.

Persistence of Automatically Tuned Values

Oracle Database remembers the sizes of the automatically tuned components across instance shutdowns if you are using a server parameter file (SPFILE). As a result, the system does need to learn the characteristics of the workload again each time an instance is started. It can begin with information from the past instance and continue evaluating workload where it left off at the last shutdown.

Organization of the Database Buffer Cache

The buffers in the cache are organized in two lists: the write list and the least recently used (LRU) list. The write list holds dirty buffers, which contain data that has been modified but has not yet been written to disk. The LRU list holds free buffers, pinned buffers, and dirty buffers that have not yet been moved to the write list. Free buffers do not contain any useful data and are available for use. Pinned buffers are currently being accessed.

When an Oracle process accesses a buffer, the process moves the buffer to the most recently used (MRU) end of the LRU list. As more buffers are continually moved to the MRU end of the LRU list, dirty buffers age toward the LRU end of the LRU list.

The first time an Oracle user process requires a particular piece of data, it searches for the data in the database buffer cache. If the process finds the data already in the cache (a cache hit), it can read the data directly from memory. If the process cannot find the data in the cache (a cache miss), it must copy the data block from a datafile on disk into a buffer in the cache before accessing the data. Accessing data through a cache hit is faster than data access through a cache miss.

Before reading a data block into the cache, the process must first find a free buffer. The process searches the LRU list, starting at the least recently used end of the list. The process searches either until it finds a free buffer or until it has searched the threshold limit of buffers.

If the user process finds a dirty buffer as it searches the LRU list, it moves that buffer to the write list and continues to search. When the process finds a free buffer, it reads the data block from disk into the buffer and moves the buffer to the MRU end of the LRU list.

If an Oracle user process searches the threshold limit of buffers without finding a free buffer, the process stops searching the LRU list and signals the DBW0 background process to write some of the dirty buffers to disk.

The LRU Algorithm and Full Table Scans

When the user process is performing a full table scan, it reads the blocks of the table into buffers and puts them on the LRU end (instead of the MRU end) of the LRU list. This is because a fully scanned table usually is needed only briefly, so the blocks should be moved out quickly to leave more frequently used blocks in the cache.

You can control this default behavior of blocks involved in table scans on a table-by-table basis. To specify that blocks of the table are to be placed at the MRU end of the list during a full table scan, use the CACHE clause when creating or altering a table or cluster. You can specify this behavior for small lookup tables or large static historical tables to avoid I/O on subsequent accesses of the table.

Size of the Database Buffer Cache

Oracle supports multiple block sizes in a database. The standard block size is used for the SYSTEM tablespace. You specify the standard block size by setting the initialization parameter DB_BLOCK_SIZE. Legitimate values are from 2K to 32K.

Optionally, you can also set the size for two additional buffer pools, KEEP and RECYCLE, by setting DB_KEEP_CACHE_SIZE and DB_RECYCLE_CACHE_SIZE. These three parameters are independent of one another.

The sizes and numbers of non-standard block size buffers are specified by the following parameters:

DB_2K_CACHE_SIZE
DB_4K_CACHE_SIZE
DB_8K_CACHE_SIZE
DB_16K_CACHE_SIZE
DB_32K_CACHE_SIZE

Each parameter specifies the size of the cache for the corresponding block size.

DB_BLOCK_SIZE=4096
DB_CACHE_SIZE=1024M
DB_2K_CACHE_SIZE=256M
DB_8K_CACHE_SIZE=512M

Multiple Buffer Pools

You can configure the database buffer cache with separate buffer pools that either keep data in the buffer cache or make the buffers available for new data immediately after using the data blocks. Particular schema objects (tables, clusters, indexes, and partitions) can then be assigned to the appropriate buffer pool to control the way their data blocks age out of the cache.

• The KEEP buffer pool retains the schema object's data blocks in memory.

• The RECYCLE buffer pool eliminates data blocks from memory as soon as they are no longer needed.

• The DEFAULT buffer pool contains data blocks from schema objects that are not assigned to any buffer pool, as well as schema objects that are explicitly assigned to the DEFAULT pool.

The initialization parameters that configure the KEEP and RECYCLE buffer pools are DB_KEEP_CACHE_SIZE and DB_RECYCLE_CACHE_SIZE.

Library Cache

The library cache includes the shared SQL areas, private SQL areas (in the case of a shared server configuration), PL/SQL procedures and packages, and control structures such as locks and library cache handles.

Shared SQL areas are accessible to all users, so the library cache is contained in the shared pool within the SGA.

Shared SQL Areas and Private SQL Areas

Oracle represents each SQL statement it runs with a shared SQL area and a private SQL area. Oracle recognizes when two users are executing the same SQL statement and reuses the shared SQL area for those users. However, each user must have a separate copy of the statement's private SQL area.

PL/SQL Program Units and the Shared Pool

Oracle processes PL/SQL program units (procedures, functions, packages, anonymous blocks, and database triggers) much the same way it processes individual SQL statements. Oracle allocates a shared area to hold the parsed, compiled form of a program unit. Oracle allocates a private area to hold values specific to the session that runs the program unit, including local, global, and package variables (also known as package instantiation) and buffers for executing SQL. If more than one user runs the same program unit, then a single, shared area is used by all users, while each user maintains a separate copy of his or her private SQL area, holding values specific to his or her session.

Individual SQL statements contained within a PL/SQL program unit are processed as described in the previous sections. Despite their origins within a PL/SQL program unit, these SQL statements use a shared area to hold their parsed representations and a private area for each session that runs the statement.

Dictionary Cache

The data dictionary is a collection of database tables and views containing reference information about the database, its structures, and its users. Oracle accesses the data dictionary frequently during SQL statement parsing. This access is essential to the continuing operation of Oracle.

The data dictionary is accessed so often by Oracle that two special locations in memory are designated to hold dictionary data. One area is called the data dictionary cache, also known as the row cache because it holds data as rows instead of buffers (which hold entire blocks of data). The other area in memory to hold dictionary data is the library cache. All Oracle user processes share these two caches for access to data dictionary information.

Allocation and Reuse of Memory in the Shared Pool

In general, any item (shared SQL area or dictionary row) in the shared pool remains until it is flushed according to a modified LRU algorithm. The memory for items that are not being used regularly is freed if space is required for new items that must be allocated some space in the shared pool. A modified LRU algorithm allows shared pool items that are used by many sessions to remain in memory as long as they are useful, even if the process that originally created the item terminates. As a result, the overhead and processing of SQL statements associated with a multiuser Oracle system is minimized.

When a SQL statement is submitted to Oracle for execution, Oracle automatically performs the following memory allocation steps:

1. Oracle checks the shared pool to see if a shared SQL area already exists for an identical statement. If so, that shared SQL area is used for the execution of the subsequent new instances of the statement. Alternatively, if there is no shared SQL area for a statement, Oracle allocates a new shared SQL area in the shared pool. In either case, the user's private SQL area is associated with the shared SQL area that contains the statement.

   Note:

   A shared SQL area can be flushed from the shared pool, even if the shared SQL area corresponds to an open cursor that has not been used for some time. If the open cursor is subsequently used to run its statement, Oracle reparses the statement, and a new shared SQL area is allocated in the shared pool.

2. Oracle allocates a private SQL area on behalf of the session. The location of the private SQL area depends on the type of connection established for the session.

Oracle also flushes a shared SQL area from the shared pool in these circumstances:

• When the ANALYZE statement is used to update or delete the statistics of a table, cluster, or index, all shared SQL areas that contain statements referencing the analyzed schema object are flushed from the shared pool. The next time a flushed statement is run, the statement is parsed in a new shared SQL area to reflect the new statistics for the schema object.

• If a schema object is referenced in a SQL statement and that object is later modified in any way, the shared SQL area is invalidated (marked invalid), and the statement must be reparsed the next time it is run.

• If you change a database's global database name, all information is flushed from the shared pool.

• The administrator can manually flush all information in the shared pool to assess the performance (with respect to the shared pool, not the data buffer cache) that can be expected after instance startup without shutting down the current instance. The statement ALTER SYSTEM FLUSH SHARED_POOL is used to do this.

Physical Memory

The LOCK_SGA parameter locks the SGA into physical memory.

SGA Starting Address

The SHARED_MEMORY_ADDRESS and HI_SHARED_MEMORY_ADDRESS parameters specify the SGA's starting address at runtime. These parameters are rarely used. For 64-bit platforms, HI_SHARED_MEMORY_ADDRESS specifies the high order 32 bits of the 64-bit address.

Extended Buffer Cache Mechanism

The USE_INDIRECT_DATA_BUFFERS parameter enables the use of the extended buffer cache mechanism for 32-bit platforms that can support more than 4 GB of physical memory. On platforms that do not support this much physical memory, this parameter is ignored.

Private SQL Area

A private SQL area contains data such as bind information and runtime memory structures. Each session that issues a SQL statement has a private SQL area. Each user that submits the same SQL statement has his or her own private SQL area that uses a single shared SQL area. Thus, many private SQL areas can be associated with the same shared SQL area.

The private SQL area of a cursor is itself divided into two areas whose lifetimes are different:

• The persistent area, which contains, for example, bind information. It is freed only when the cursor is closed.

• The run-time area, which is freed when the execution is terminated.

Oracle creates the runtime area as the first step of an execute request. For INSERT, UPDATE, and DELETE statements, Oracle frees the runtime area after the statement has been run. For queries, Oracle frees the runtime area only after all rows are fetched or the query is canceled.

The location of a private SQL area depends on the type of connection established for a session. If a session is connected through a dedicated server, private SQL areas are located in the server process's PGA. However, if a session is connected through a shared server, part of the private SQL area is kept in the SGA.

Session Memory

Session memory is the memory allocated to hold a session's variables (logon information) and other information related to the session. For a shared server, the session memory is shared and not private.