外文翻譯---解析oracle數(shù)據(jù)庫(kù)后臺(tái)進(jìn)程的功能

1、<p>　　耿祥義.JSP 基礎(chǔ)教程.北京:清華大學(xué)出版社,2009.10</p><p><b>　　附錄</b></p><p><b>　　英文原文</b></p><p>　　Analysis of Oracle database background process function</p>

2、<p>　　1.DBWR process: the process of the implementation of the buffer is written to the data file, is responsible for the buffer memory management in a Oracle background process. When a buffer buffer is modified,

3、it is marked as" dirty", DBWR is the main task of" dirty" buffer is written to disk, so that the buffer to maintain" clean". Because the buffer memory buffer in the database or user process

4、dirty, unused buffer to reduce the number of. When the unused buffer down to little, so that user proce</p><p>　　Oracle uses LRU ( LEAST RECENTLY USED ) algorithm ( LRU ) keep the memory data block is a rece

5、ntly used, so that the I/O minimum. In the following case indicates that DBWR will dirty buffers to disk:</p><p>　　When a server process will be a buffer into the" dirty" table, the dirty expressio

6、n to the critical length, the service will notify DBWR writing. The critical length for the value of the parameter DB-BLOCK-WRITE-BATCH half.</p><p>　　When a server process in a LRU look-up table in DB-BLOCK

7、-MAX-SCAN-CNT buffer, did not find unused buffer, it stops to find and notify the DBWR to write. Timeout ( every3 seconds ), DBWR will inform itself. When a checkpoint, LGWR will inform DBWR. of the first two cases, DBWR

8、 will dirty table block is written to disk, each can write blocks by the initialization parameter DB-BLOCK- specified by WRITE-BATCH. If the dirty list without this parameter specifies the number of buffer, DBWR from the

9、 LUR</p><p>　　If the DBWR in three inactive, a time-out occurs. In this case the DBWR on the LRU table to find the specified number of buffer, to find any dirty buffers to disk. Whenever a timeout occurs, DB

10、WR finds a new buffer group. Each time by the DBWR to find the number of buffers for sleep parameter DB-BLOCK- WRITE-BATCH values two times. If the database is idle, DBWR will eventually all buffer storage to disk.</p

11、><p>　　In the occurs check point, LGWR specifies a modified buffer table must be written to disk. DBWR the specified buffer is written to disk.</p><p>　　On some platforms, one instance can have mul

12、tiple DBWR. in this instance, some blocks can be written to a disk, some other piece can be written to other disk. The parameter DB-WRITERS control of DBWR process number.</p><p>　　2. LGWR process: the proce

13、ss will log buffer is written to disk in a log file, it is responsible for the management of log buffer of a Oracle background process. LGWR process will since the last written to disk since all log entries for output, t

14、he output of the LGWR:</p><p>　　◆When a user process to submit a written when a commit record transaction.</p><p>　　◆Every three seconds will log buffer output.</p><p>　　◆When the l

15、og buffer1/3 is full when the log buffer output.</p><p>　　◆When DBWR will modify the buffer is written to disk when the log buffer output.</p><p>　　The LGWR process synchronous writes to the act

16、ive mirror online log file group. If the group a file to be deleted or not available, LGWR can continue to write to the group and other documents.</p><p>　　Log buffer is a circular buffer. When the LGWR log

17、buffer log entries are written to the log file, the server process can be a new log entry is written to the log buffer. LGWR used to write very quickly, can ensure that the log buffer total space can be written into the

18、new log entries.</p><p>　　Note: sometimes when we need more log buffer, LWGR in a transaction log entries before will write log entries, and these only when in the future after the transaction commits to per

19、manent.</p><p>　　ORACLE uses fast delivery mechanism, when the user issues a COMMIT statement, an COMMIT record immediately into the log buffer, but the corresponding data buffer change is to be delayed, unt

20、il more effective only when they are written to the data file. When a transaction is committed, is assigned to a system change number (SCN ), which together with the transaction log records in the log. Because the SCN re

21、corded in the log, so that in the parallel server configuration, recovery operations can b</p><p>　　3. CKPT process: the process at the inspection point appears, all the data file headers are modified, indic

22、ating the check point. Usually, the task performed by LGWR. However, if the check point significantly degrade the system performance, the CKPT process running, from the original LGWR process of the implementation of the

23、checkpoint job separated, by the CKPT process to achieve. For many applications, the CKPT process is not necessary. Only when the database has many data files, LGWR in check</p><p>　　The initialization param

24、eter CHECKPOINT-PROCESS control of CKPT process that can or not. The default for FALSE, namely, that can't.</p><p>　　4. SMON process: the process instance startup, perform instance recovery, is also resp

25、onsible for cleaning up no longer use temporary segments. In a parallel server environment, SMON on fault CPU or instance instance recovery. The SMON process is regularly called up, check is needed, or other process that

26、 needs to be invoked.</p><p>　　5. PMON process: the process in a user process failure recovery implementation process, responsible for the cleaning of memory storage area and release the resources used by th

27、e process. Example: it is to reset the events table status, releasing the blockade, the failure process of ID from activity in the process table remove. PMON also periodically checks the scheduling process ( DISPATCHER )

28、 and a server process state, if the dead, then restart ( not including intentional deletion process ).</p><p>　　PMON regularly called up, check is needed, or other process that needs to be invoked.</p>

29、<p>　　6. RECO process: the process is in a distributed option when used by a process, automatic solution in the distributed transaction failure. A node RECO background process automatically connect to include othe

30、r be in suspense distributed transaction database, RECO automatically solve all pending and affairs. Any corresponding to the processed is pending transaction is one from each database table delete suspended matters.<

31、/p><p>　　When a database server RECO background process attempts to establish a remote server, if the remote server is not available or network connection cannot be established, RECO automatically at a time int

32、erval after connect again.</p><p>　　RECO background process only when allows distributed transaction system, and DISTRIBUTED C TRANSACTIONS parameter is greater than 0.</p><p>　　7. ARCH process:

33、 the process will have to fill the online log files are copied to the specified storage device. When the log is ARCHIVELOG use, and can automatically archiving ARCH process exists.</p><p>　　8. The LCKn proce

34、ss is: with parallel server option environment, can be up to10 processes ( LCK0, LCK1... ... For example, LCK9), the block between the.</p><p>　　9. ( Dnnn process scheduling process ): the process allows the

35、 user process shared limited server process ( SERVER PROCESS ). No scheduling process, each user process requires a dedicated server process ( DEDICATEDSERVER PROCESS ). Multi-threaded server ( MULTI-THREADED SERVER ) ca

36、n support multiple user processes. If the system has a large number of users, many cues server can support a large users, especially in _ client server environment.</p><p>　　In a database instance can set up

37、 multiple scheduling process. For each kind of network protocol to establish at least one dispatcher processes. Database administrator according to the operating system of each process can be connected with restrictions

38、on the number of decisions to initiate the scheduler optimal number, when the instance is running can be added to or deleted from the scheduling process. Multiple cue server requires SQL*NET version 2 or later version. I

39、n a multiple cue server con</p><p><b>　　中文譯文</b></p><p>　　解析Oracle數(shù)據(jù)庫(kù)后臺(tái)進(jìn)程的功能</p><p>　　1、DBWR進(jìn)程：該進(jìn)程執(zhí)行將緩沖區(qū)寫(xiě)入數(shù)據(jù)文件，是負(fù)責(zé)緩沖存儲(chǔ)區(qū)管理的一個(gè)Oracle后臺(tái)進(jìn)程。當(dāng)緩沖區(qū)中的一緩沖區(qū)被修改，它被標(biāo)志為“弄臟”，DBWR的主要任務(wù)是

40、將“弄臟”的緩沖區(qū)寫(xiě)入磁盤(pán)，使緩沖區(qū)保持“干凈”。由于緩沖存儲(chǔ)區(qū)的緩沖區(qū)填入數(shù)據(jù)庫(kù)或被用戶(hù)進(jìn)程弄臟，未用的緩沖區(qū)的數(shù)目減少。當(dāng)未用的緩沖區(qū)下降到很少，以致用戶(hù)進(jìn)程要從磁盤(pán)讀入塊到內(nèi)存存儲(chǔ)區(qū)時(shí)無(wú)法找到未用的緩沖區(qū)時(shí)，DBWR將管理緩沖存儲(chǔ)區(qū)，使用戶(hù)進(jìn)程總可得到未用的緩沖區(qū)。 </p><p>　　Oracle采用LRU（LEAST RECENTLY USED）算法（最近最少使用算法）保持內(nèi)存中的數(shù)據(jù)塊是最近使用的，

41、使I/O最小。在下列情況預(yù)示DBWR 要將弄臟的緩沖區(qū)寫(xiě)入磁盤(pán)： </p><p>　　當(dāng)一個(gè)服務(wù)器進(jìn)程將一緩沖區(qū)移入“弄臟”表，該弄臟表達(dá)到臨界長(zhǎng)度時(shí)，該服務(wù)進(jìn)程將通知DBWR進(jìn)行寫(xiě)。該臨界長(zhǎng)度是為參數(shù)DB-BLOCK-WRITE-BATCH的值的一半。 </p><p>　　當(dāng)一個(gè)服務(wù)器進(jìn)程在LRU表中查找DB-BLOCK-MAX-SCAN-CNT緩沖區(qū)時(shí)，沒(méi)有查到未用的緩沖區(qū)，它停止

42、查找并通知DBWR進(jìn)行寫(xiě)。出現(xiàn)超時(shí)（每次3秒），DBWR 將通知本身。當(dāng)出現(xiàn)檢查點(diǎn)時(shí)，LGWR將通知DBWR.在前兩種情況下，DBWR將弄臟表中的塊寫(xiě)入磁盤(pán)，每次可寫(xiě)的塊數(shù)由初始化參數(shù)DB-BLOCK- WRITE-BATCH所指定。如果弄臟表中沒(méi)有該參數(shù)指定塊數(shù)的緩沖區(qū)，DBWR從LUR表中查找另外一個(gè)弄臟緩沖區(qū)。 </p><p>　　如果DBWR在三秒內(nèi)未活動(dòng)，則出現(xiàn)超時(shí)。在這種情況下DBWR對(duì)LRU表查找

43、指定數(shù)目的緩沖區(qū)，將所找到任何弄臟緩沖區(qū)寫(xiě)入磁盤(pán)。每當(dāng)出現(xiàn)超時(shí)，DBWR查找一個(gè)新的緩沖區(qū)組。每次由DBWR查找的緩沖區(qū)的數(shù)目是為寢化參數(shù)DB-BLOCK- WRITE-BATCH的值的二倍。如果數(shù)據(jù)庫(kù)空運(yùn)轉(zhuǎn)，DBWR最終將全部緩沖區(qū)存儲(chǔ)區(qū)寫(xiě)入磁盤(pán)。 </p><p>　　在出現(xiàn)檢查點(diǎn)時(shí)，LGWR指定一修改緩沖區(qū)表必須寫(xiě)入到磁盤(pán)。DBWR將指定的緩沖區(qū)寫(xiě)入磁盤(pán)。 </p><p>　　在有

44、些平臺(tái)上，一個(gè)實(shí)例可有多個(gè)DBWR.在這樣的實(shí)例中，一些塊可寫(xiě)入一磁盤(pán)，另一些塊可寫(xiě)入其它磁盤(pán)。參數(shù)DB-WRITERS控制DBWR進(jìn)程個(gè)數(shù)。 </p><p>　　2、LGWR進(jìn)程：該進(jìn)程將日志緩沖區(qū)寫(xiě)入磁盤(pán)上的一個(gè)日志文件，它是負(fù)責(zé)管理日志緩沖區(qū)的一個(gè)Oracle后臺(tái)進(jìn)程。LGWR進(jìn)程將自上次寫(xiě)入磁盤(pán)以來(lái)的全部日志項(xiàng)輸出，LGWR輸出： </p><p>　　◆當(dāng)用戶(hù)進(jìn)程提交一事務(wù)時(shí)寫(xiě)

45、入一個(gè)提交記錄。 </p><p>　　◆每三秒將日志緩沖區(qū)輸出。 </p><p>　　◆當(dāng)日志緩沖區(qū)的1/3已滿(mǎn)時(shí)將日志緩沖區(qū)輸出。 </p><p>　　◆當(dāng)DBWR將修改緩沖區(qū)寫(xiě)入磁盤(pán)時(shí)則將日志緩沖區(qū)輸出。</p><p>　　LGWR進(jìn)程同步地寫(xiě)入到活動(dòng)的鏡象在線日志文件組。如果組中一個(gè)文件被刪除或不可用，LGWR可繼續(xù)地寫(xiě)入該組的

46、其它文件。 </p><p>　　日志緩沖區(qū)是一個(gè)循環(huán)緩沖區(qū)。當(dāng)LGWR將日志緩沖區(qū)的日志項(xiàng)寫(xiě)入日志文件后，服務(wù)器進(jìn)程可將新的日志項(xiàng)寫(xiě)入到該日志緩沖區(qū)。LGWR 通常寫(xiě)得很快，可確保日志緩沖區(qū)總有空間可寫(xiě)入新的日志項(xiàng)。 </p><p>　　注意：有時(shí)候當(dāng)需要更多的日志緩沖區(qū)時(shí)，LWGR在一個(gè)事務(wù)提交前就將日志項(xiàng)寫(xiě)出，而這些日志項(xiàng)僅當(dāng)在以后事務(wù)提交后才永久化。 </p>&l

47、t;p>　　ORACLE使用快速提交機(jī)制，當(dāng)用戶(hù)發(fā)出COMMIT語(yǔ)句時(shí)，一個(gè)COMMIT記錄立即放入日志緩沖區(qū)，但相應(yīng)的數(shù)據(jù)緩沖區(qū)改變是被延遲，直到在更有效時(shí)才將它們寫(xiě)入數(shù)據(jù)文件。當(dāng)一事務(wù)提交時(shí)，被賦給一個(gè)系統(tǒng)修改號(hào)（SCN），它同事務(wù)日志項(xiàng)一起記錄在日志中。由于SCN記錄在日志中，以致在并行服務(wù)器選項(xiàng)配置情況下，恢復(fù)操作可以同步。 </p><p>　　3、CKPT進(jìn)程：該進(jìn)程在檢查點(diǎn)出現(xiàn)時(shí)，對(duì)全部數(shù)據(jù)文

48、件的標(biāo)題進(jìn)行修改，指示該檢查點(diǎn)。在通常的情況下，該任務(wù)由LGWR執(zhí)行。然而，如果檢查點(diǎn)明顯地降低系統(tǒng)性能時(shí)，可使CKPT進(jìn)程運(yùn)行，將原來(lái)由LGWR進(jìn)程執(zhí)行的檢查點(diǎn)的工作分離出來(lái)，由CKPT進(jìn)程實(shí)現(xiàn)。對(duì)于許多應(yīng)用情況，CKPT進(jìn)程是不必要的。只有當(dāng)數(shù)據(jù)庫(kù)有許多數(shù)據(jù)文件，LGWR在檢查點(diǎn)時(shí)明顯地降低性能才使CKPT運(yùn)行。 CKPT進(jìn)程不將塊寫(xiě)入磁盤(pán)，該工作是由DBWR完成的。</p><p>　　初始化參數(shù)CHECK

49、POINT-PROCESS控制CKPT進(jìn)程的使能或使不能。缺省時(shí)為FALSE，即為使不能。 </p><p>　　4、SMON進(jìn)程：該進(jìn)程實(shí)例啟動(dòng)時(shí)，執(zhí)行實(shí)例恢復(fù)，還負(fù)責(zé)清理不再使用的臨時(shí)段。在具有并行服務(wù)器選項(xiàng)的環(huán)境下，SMON對(duì)有故障CPU或?qū)嵗M(jìn)行實(shí)例恢復(fù)。SMON進(jìn)程有規(guī)律地被呼醒，檢查是否需要，或者其它進(jìn)程發(fā)現(xiàn)需要時(shí)可以被調(diào)用。 </p><p>　　5、PMON進(jìn)程：該進(jìn)程在用

50、戶(hù)進(jìn)程出現(xiàn)故障時(shí)執(zhí)行進(jìn)程恢復(fù)，負(fù)責(zé)清理內(nèi)存儲(chǔ)區(qū)和釋放該進(jìn)程所使用的資源。例：它要重置活動(dòng)事務(wù)表的狀態(tài)，釋放封鎖，將該故障的進(jìn)程的ID從活動(dòng)進(jìn)程表中移去。PMON還周期地檢查調(diào)度進(jìn)程（DISPATCHER）和服務(wù)器進(jìn)程的狀態(tài)，如果已死，則重新啟動(dòng)（不包括有意刪除的進(jìn)程）。 </p><p>　　6、RECO進(jìn)程：該進(jìn)程是在具有分布式選項(xiàng)時(shí)所使用的一個(gè)進(jìn)程，自動(dòng)地解決在分布式事務(wù)中的故障。一個(gè)結(jié)點(diǎn)RECO后臺(tái)進(jìn)程自動(dòng)

51、地連接到包含有懸而未決的分布式事務(wù)的其它數(shù)據(jù)庫(kù)中，RECO自動(dòng)地解決所有的懸而不決的事務(wù)。任何相應(yīng)于已處理的懸而不決的事務(wù)的行將從每一個(gè)數(shù)據(jù)庫(kù)的懸掛事務(wù)表中刪去。 </p><p>　　當(dāng)一數(shù)據(jù)庫(kù)服務(wù)器的RECO后臺(tái)進(jìn)程試圖建立同一遠(yuǎn)程服務(wù)器的通信，如果遠(yuǎn)程服務(wù)器是不可用或者網(wǎng)絡(luò)連接不能建立時(shí)，RECO自動(dòng)地在一個(gè)時(shí)間間隔之后再次連接。 </p><p>　　RECO后臺(tái)進(jìn)程僅當(dāng)在允許分布

52、式事務(wù)的系統(tǒng)中出現(xiàn)，而且DISTRIBUTED C TRANSACTIONS參數(shù)是大于0。 </p><p>　　7、ARCH進(jìn)程：該進(jìn)程將已填滿(mǎn)的在線日志文件拷貝到指定的存儲(chǔ)設(shè)備。當(dāng)日志是為ARCHIVELOG使用方式、并可自動(dòng)地歸檔時(shí)ARCH進(jìn)程才存在。 </p><p>　　8、LCKn進(jìn)程：是在具有并行服務(wù)器選件環(huán)境下使用，可多至10個(gè)進(jìn)程（LCK0，LCK1……，LCK9），用于

53、實(shí)例間的封鎖。 </p><p>　　9、Dnnn進(jìn)程（調(diào)度進(jìn)程）：該進(jìn)程允許用戶(hù)進(jìn)程共享有限的服務(wù)器進(jìn)程（SERVER PROCESS）。沒(méi)有調(diào)度進(jìn)程時(shí)，每個(gè)用戶(hù)進(jìn)程需要一個(gè)專(zhuān)用服務(wù)進(jìn)程（DEDICATEDSERVER PROCESS）。對(duì)于多線索服務(wù)器（MULTI-THREADED SERVER）可支持多個(gè)用戶(hù)進(jìn)程。如果在系統(tǒng)中具有大量用戶(hù)，多線索服務(wù)器可支持大量用戶(hù)，尤其在客戶(hù)_服務(wù)器環(huán)境中。 </p

54、><p>　　在一個(gè)數(shù)據(jù)庫(kù)實(shí)例中可建立多個(gè)調(diào)度進(jìn)程。對(duì)每種網(wǎng)絡(luò)協(xié)議至少建立一個(gè)調(diào)度進(jìn)程。數(shù)據(jù)庫(kù)管理員根據(jù)操作系統(tǒng)中每個(gè)進(jìn)程可連接數(shù)目的限制決定啟動(dòng)的調(diào)度程序的最優(yōu)數(shù)，在實(shí)例運(yùn)行時(shí)可增加或刪除調(diào)度進(jìn)程。多線索服務(wù)器需要SQL*NET版本2或更后的版本。在多線索服務(wù)器的配置下，一個(gè)網(wǎng)絡(luò)接收器進(jìn)程等待客戶(hù)應(yīng)用連接請(qǐng)求，并將每一個(gè)發(fā)送到一個(gè)調(diào)度進(jìn)程。如果不能將客戶(hù)應(yīng)用連接到一調(diào)度進(jìn)程時(shí)，網(wǎng)絡(luò)接收器進(jìn)程將啟動(dòng)一個(gè)專(zhuān)用服務(wù)器進(jìn)程

55、。該網(wǎng)絡(luò)接收器進(jìn)程不是Oracle實(shí)例的組成部分，它是處理與Oracle有關(guān)的網(wǎng)絡(luò)進(jìn)程的組成部分。在實(shí)例啟動(dòng)時(shí)，該網(wǎng)絡(luò)接收器被打開(kāi)，為用戶(hù)連接到Oracle建立一通信路徑，然后每一個(gè)調(diào)度進(jìn)程把連接請(qǐng)求的調(diào)度進(jìn)程的地址給予它的接收器。當(dāng)一個(gè)用戶(hù)進(jìn)程作連接請(qǐng)求時(shí)，網(wǎng)絡(luò)接收器進(jìn)程分析請(qǐng)求并決定該用戶(hù)是否可使用一調(diào)度進(jìn)程。如果是，該網(wǎng)絡(luò)接收器進(jìn)程返回該調(diào)度進(jìn)程的地址，之后用戶(hù)進(jìn)程直接連接到該調(diào)度進(jìn)程。有些用戶(hù)進(jìn)程不能調(diào)度進(jìn)程通信（如果使用SQL