ORACLE instance of the process of the structure and memory structure

Apr 9 posted in Oracle (RSS), 160 views.

ORACLE instance of the process of the structure and memory structure

1. ORACLE instance of the process of the structure and memory structure
1) the process of the structure of the process is a mechanism for the operating system, which operated a series of executable steps. In some operating systems used in operations (JOB) or task (TASK) terminology. A process normally has its own dedicated storage area. ORACL
E architecture design process so that the greatest performance.
ORACLE instance, there are two types: single-process and multi-process instance of an example.
Single process ORACLE (also known as single-user ORACLE) is a database system, a process to implement all the ORAC
LE code. ORACLE part as a result of applications and customers can not separate the process of implementation, so the code ORACLE database applications and users is the implementation of a single process.
Process in a single instance of ORACLE environment, to allow only one user can access. For example, in MS-DOS running on O
RACLE.
ORACLE instance of multi-process (also known as multi-user ORACLE) using multiple processes to execute different parts of ORACLE, a connection for each user has a process.
In a multi-process system, the process is divided into two categories: the process of the user process and ORACLE. When a user runs an application, such as the PRO * C program or an ORACLE tool (such as SQL * PLUS), for the user running the application process to establish a user. ORACLE process is divided into two categories: server processes and background processes. Server process for processing connected to the instance the request of the user process. When the application and ORACELE is running on the same machine,
Rather than through the network, the user process in general and its corresponding server process into a single process, can reduce system overhead. However, when the application and ORACLE to run on different machines, users go through a separation process with the ORACLE server process communication. It executable the following tasks:
 issued on the application of the SQL statement for syntax analysis and implementation.
 from the disk (data files) to read the necessary data blocks to the shared database buffers of the SGA (
When the block is not the buffer zone).
 the results back to the application procedures.
In order to make the best system performance and coordination of multiple users in a multi-process system, the use of some additional process, known as background processes. In many operating systems, the background process is automatically in the instance to create. An ORAC
LE can be many examples of the background process, but they are not persistent. The name of the background process as follows:
DBWR Database Writer
LGWR Log Writer
CKPT checkpoint
SMON System Monitor
PMON process monitoring
ARCH archive
The restoration of RECO
LCKn blockade
Dnnn scheduling process
Snnn server and ORACLE background processes for each of the different parts of the database interaction.
The following features of the background process for a brief introduction:
DBWR process: the process of implementation of the data files will be written into the buffer zone, buffer storage area is responsible for the management of an OR
ACLE background processes. When the buffer of a buffer is modified, it is marked as "dirty", DBWR will be the main task is to "dirty" disk write buffer to the buffer zone to maintain "clean." As the buffer storage areas fill in the database buffer dirty or user process, not by a decrease in the number of the buffer zone. When not used in the buffer zone fell to very few, so that the user process block read from disk into memory storage area can not be found when the buffer is not used, DBWR will manage the buffer storage area, so that users can not use the total process of buffer.
ORACLE using LRU (LEAST RECENTLY USED) algorithm (least recently used algorithm) to maintain the data in memory is the most recently used block, so that I / O minimum. Foreshadowed in the following circumstances DBWR write to the disk buffer dirty:
 When a server process will be a buffer zone into the "dirty" table, the expression of the critical length dirty, the service will notify the DBWR process to write. The critical parameter is the length of DB-BLOCK-WRITE-BATCH
Half the value.
 When a server process in the LRU table to find DB-BLOCK-MAX-SCAN-CNT buffer, no buffer available unused, it ceased to find and notify the DBWR to write.
 emerged overtime (each 3 seconds), DBWR will inform itself.
 When the check point, LGWR will notify DBWR
The former two cases, DBWR dirty table will be written into the disk blocks, each block can be written from initialization parameters
DB-BLOCK-WRITE-BATCH specified. If you do not dirty the table parameter specifies the number of buffer blocks, DB
WR table from the LUR find another dirty buffer.
If not DBWR activity in three seconds, while there is overtime. In this case DBWR to specify the number of LRU table to find the buffer zone, will have to find any dirty buffers written to disk. Whenever there is overtime, DBWR buffer to find a new group. Find each DBWR buffer by the number of sleep parameters for the DB-BLOCK-WRITE-BAT
Twice the value of CH. Empty if the database is running, DBWR final storage area will be written into the disk buffer.
In the event of check points, LGWR modified to designate a buffer zone required to be written to disk. DBWR will write to the disk buffer specified.
In some platforms, an example may have multiple DBWR. In such cases, some can write a disk block, and some other block can be written into the disk. DB-WRITERS control parameters of the process of the number of DBWR.
LGWR process: the process will be written into the log buffer of a disk log file, which is responsible for managing a log buffer ORACLE background processes. LGWR process will write to the disk since the last full year of output log, LGWR output:
 When the user when the process of submitting a written record of a submission.
 log every three seconds to buffer the output.
 When the log buffer of 1 / 3 full when the log output buffer.
 When the DBWR write to the disk buffer zone will be changed when the log buffer will be output.
LGWR process synchronously written to mirror the activities of online log file groups. If the group a file is deleted or unavailable, LGWR can continue to write other documents of the group.
Log buffer is a buffer cycle. When the log buffer LGWR will write to the log of the log file,
New server process may be written to the log of the log buffer. LGWR usually write very quickly to ensure that there will always be room for the log buffer to log into the new item.
Note: Sometimes when the need for more when the log buffer, LWGR services in a log before items will be written, and these only when the log of affairs submitted at a later date after the permanent.
ORACLE submitted using the fast mechanism, when a user issued COMMIT statement, a COMMIT log record buffer Add immediately, but the corresponding change in the data buffer is delayed until more effective when they are written into the data files. When submitted to a Panel, they were assigned to a system changes its (SCN), which together with the transaction logs recorded in the log. Since the SCN recorded in the log, resulting in parallel server configuration options, the recovery operations can take place in parallel.

CKPT process: the process in the checkpoint occurs, all data on the title of the document be amended to instruct the checkpoint. In normal circumstances, the task of implementation by the LGWR. However, if the checkpoint is clear that reducing system performance, the CKPT process can run the original LGWR process from the implementation of the work of the inspection points separated
To achieve by the CKPT process. For many applications, CKPT process is unnecessary. Database only when there are many data files, LGWR when the checkpoint is obviously able to reduce the performance to run CKPT. CKPT process does not write to disk blocks, the work is completed by the DBWR.
CHECKPOINT-PROCESS initialization parameter control to enable the CKPT process or not. When the default FALSE,
Is not so.

SMON process: the process of implementation of examples of examples of when to start recovery, is also responsible for cleaning up the temporary segment is no longer used. With the Parallel Server option in the circumstances, SMON failure of CPU or to resume examples examples. SMON process being called up regularly to check whether there is a need, or other process of discovery can be called when necessary.

PMON process: the process of failure of the user when the process of implementation of the process of recovery, is responsible for clean-up inside the storage area and the release of the resources used in the process. For example: it is to reset the active transaction table, and the release of the blockade, the failure of the process ID from the activities of the process of removal of the table. PMON also periodically check the scheduling process (DISPATCHER) and the server process, and if dead, then restart (not including the intention to delete the process).
PMON regularly being called up to check whether there is a need, or other process of discovery can be called when necessary.

RECO process: the process is distributed in the option of a process used to automatically resolve the distributed transaction failures. A node RECO background process automatically connect to contain unresolved distributed transaction in the other database, RECO automatically resolve all pending matters. Have been processed in any of the corresponding pending transaction row from the database for each table by deleting the hoisting of affairs.
When a database server of the RECO background process of attempting to set up the same remote server communication, if the remote server is not available or when the network connection could not be established, RECO automatically after a time interval to connect again.
RECO background process is allowed only if the distributed transaction systems appear, and DISTRIBUTED - TRANS
ACTIONS parameter is greater than 0.

ARCH process: the process will have to fill up the online log files are copied to the designated storage devices. When the log is
Use ARCHIVELOG mode and automatic archiving only exist when the ARCH process.

LCKn process: it is in a parallel server environment options can be up to 10 process (LCK0, LCK1
... ..., LCK9), for example between the blockade.

Dnnn process (scheduling process): The process allows the user to process a limited share of the server process (SERVER P
ROCESS). There is no scheduling process, each user process requires a dedicated server process (DEDICATEDS
ERVER PROCESS). For more clues to the server (MULTI-THREADED SERVER) can support multiple use
Households in the process. If in the system with a large number of users, and more clues to the server can support a large number of users, particularly in the client server environment _.
Examples in a database can create multiple scheduling process. Network protocol for each of at least the establishment of a scheduling process. Database Administrator in accordance with the process of the operating system can be connected to each restriction on the number of the scheduler decided to start a few of the best in the run-time instance of scheduling the process to increase or delete. Need more clues to the server SQL * NET version 2 or later versions. Clues in a multi-server configuration, a network process to wait for the receiver client application connection requests, and each sent to a dispatching process. If the client application can not be connected to a scheduling process, the network receiver process will start a dedicated server process. The network receiver process is not OR
An integral part of ACLE example, it is to deal with the network ORACLE process. In the example starts, the network receiver is opened, for the user to connect to ORACLE to establish a communication path, and then a scheduling process for each connection request to the scheduling process to give it the address of the receiver. When a user process requests for connections, the network analysis of the receiver process the request and decide whether or not the user can use a scheduling process. If so, the process of the network to return to the receiver's address scheduling process, the process after the user to connect directly to the scheduling process. Some users can not process scheduling process communications (if you are using SQL * NET users of previous versions),
Network receiver process so users can not connect to a scheduling process. In this case, the network receiver to establish a dedicated server process, to establish a suitable connection.

2), ORACLE memory structure
ORACLE memory storage in the following information:
 Implementation of program code.
 Connect the session information
 procedures required during the implementation of data and information sharing
 storage from the buffer storage of information.
ORACLE memory have the following basic structure:
 software code area
 overall system, including the database buffer storage area, the log buffer and shared pool.
 overall process, including the stack area and data area.
Sort District 

Area used to store software code is implemented or the implementation of the program code.
Software areas are read-only, can be installed into a shared or non-sharing. ORACLE system process can be shared, resulting in a number of
ORACLE users can access it without having to have multiple copies in memory. User program can be shared may or may not be shared.

System Global Area for a group of distribution by the ORACLE shared memory structure, a database instance can contain data or control information. If multiple users connect to the same instance at the same time, the SGA in the case of data can be shared by multiple users, it is also known as the shared global area. When the instance starts, SGA memory automatically be assigned; when the instance shut down, the store has been recovered. All connected to the multi-process database instance all users can be assigned automatically; when the instance shut down, the store has been recovered. All connected to the multi-process database instance all users can use their information in the SGA, but only a few can be written into the information process. Information stored in the SGA memory is divided into several areas: the database buffer storage area, the log buffer, shared pool, request and response queues, data dictionary storage areas and other information.

Procedures for the overall district
PGA is a memory area that contains a single process data and control information, so the overall area is also known as the process (P
ROCESS GLOBAL AREA).

Sort sort area memory space needs, ORACLE to use the memory to sort data, this part of space known as the sorting area. Sort sort area exists in the user process requests memory, the space for the appropriate amount of data on the sort of size, can be growth, but by the initialization parameter SORT-AREA-SIZER restricted.