Oracle's architecture

Just debut, here hope you Daniel guidance, hope here to become a platform for my growth! This blog I tell you about Oracle's architecture, but also a test of their own learning!

Oralce memory is composed of SGA+PGA.

The Oracle server is comprised of Oracle instances and Oracle databases. Here's a more detailed look.

Oracle instances:

The user must be able to access the Oracle database through an Oracle instance

An Oracle instance can only correspond to one Oracle database

An Oracle instance consists of a memory structure and a background process structure.

Memory structure: The SGA (System global Area) systems globals, including the shared pool,

Database buffer cache (data buffer), redo log buffer cache (data buffer)

Larger pool (large pool), Java pool, stream pond.

Background process: Smon (System Monitor), Pmon (Process Monitor), DBWN (database write process),

LGWR (log write process), ARC (archive log process), CKPT (checkpoint).

Here the Oracle architecture contains content that has been outlined in detail below:

The shared pool contains the library cache and the data dictionary cache.

The shared pool area is an area of syntax checking, semantic checking, hash calculation, parsing, and execution of SQL statements.

In the production of share pool hit rate, should reach more than 90%, so that the CPU pressure is small.

When executing the SQL statement, a number of "data dictionaries" are called at the bottom, so you design data dictionary cache.

CBO optimizer: Measure the SQL Execution plan and choose an optimal execution method. Optimizer collects information.

2.data buffer cache is read for the first time, then stored in buffer cache, second read, directly from buffer cache

Reading data from memory is certainly faster than reading on disk.

Write operation, first write in memory, fast, it thinks the memory is finished, this operation is finished, when the library is idle and then writes the memory data to the disk DBF file

Reduce I/O pressure with a large hit ratio.

Buffer Cache Storage rule: If it is full, from the cold data, find the top priority coverage with the current time.

Redo Log Buffer Cache

Write the memory first, fast, it thinks the memory writes, this operation is finished, when the library is idle and then writes the memory data to the disk's redo log file

The main database is to write the log first, and then write the database. Log is important, log sound, data is gone, can be restored through the log

Large pool storage Big data "one time to extract large data, storage in the large pool is better, storage database Buffer cache is not good, easy to cause a hit fake image"

Large Pool storage: 1, Rman backup data 2, parallel operation.

Java pool invokes Java directives on the front end.

Stream pool allocates memory space for stream applications, as part of advanced replication technology.

PGA (Process Global Area): program Globals

The program Global Zone (PGA) contains the data and control information required for a single server process, such as the SQL binding variable passed in each session, to

and control information about the session.

If the session is sorted or hashed, the memory in the PGA is also used. (including parallel and sort direct go PGA, not

Go to buffer cache);

Background process

Pmon process monitoring process, monitoring the user process is normal.

1), clean up the failed process.

2), releases all currently pending locks.

3), release the resources used by the failed process.

Smon system monitoring process, real-time monitoring of the entire Oracle status.

1), after the instance fails, automatically resumes the instance when the database is reopened.

2), tidy up the free space of the data files, combine the adjacent areas.

3), release the temporary segment that is no longer in use.

DBWR Data Write process

1), manage the data buffers, and keep the most recently used blocks in memory.

2), write the modified buffer data to the data file.

LGWR Log Write process

1), responsible for writing log data in the log buffer to the log file.

2), the system has more than one log file, the process in a circular way to write data to the file.

CKPT Checkpoint Process

1), in order to prevent instances from crashing, in order to carry out the process of instance recovery as soon as possible.

2), Checkpoint includes two kinds: full checkpoint, incremental checkpoint.

3), the full checkpoint is triggered only in the following two cases:

1, the user issues the alter system checkpoint;

2, the database is closed normally except for shutdown abort.

3. Incremental checkpoint Trigger:

1, every three seconds.

2, online log switching.

This article from "Newbie" blog, reproduced please contact the author!

Oracle's architecture