提高ORACLE資料庫的查詢統計速度

大型資料庫系統中往往要用到查詢統計，但是對於資料量大的系統，使用者在進行複雜的查詢統計時往往感到速度很慢，不能滿足應用要求，這就要求我們在設計資料庫系統時進行合理設定，提高查詢統計的速度。本文結合筆者的項目開發經驗，闡述具體的設定方法。

以oracle7.33資料庫系統為例，我們在開發大型oracle資料庫系統時結合項目的特點，本著安全、高效的原則對資料庫進行了一些實體設計，從而大大提高了資料庫的查詢統計速度。總結為如下幾點：

1）擴大資料資料表空間到500M，用於存放本系統的資料；

2）段盤區的初始大小為10K，增長大小為10K，增長幅度為1；

3）使用者臨時空間增大40M；

4）系統暫存資料表空間和復原段資料表空間增大40M，並且建立4個復原段；

5）需要經常連接查詢，而且資料量又大的庫存表、名錄表、收發料表放在一簇內；

6）提供定時備份，備份檔案放在另外的機器上。

設定資料資料表空間的SQL語句如下：

CREATE TABLESPACE WXGL_DATA1 DATAFILE 'WXGL_DATA1.ORA' SIZE 500M ONLINE;

增加系統暫存資料表空間和復原段資料表空間的SQL語句如下：

ALTER TABLESPACE TEMPORARY_DATA ADD DATAFILE 'TMP2ORCL.ORA' SIZE 40M;ALTER TABLESPACE ROLLBACK_DATA ADD DATAFILE 'RBS2ORCL.ORA' SIZE 40M;

將資料空間設定在指定的資料檔案的SQL語句如下：

CREATE USER ZBGL IDENTIFIED BY ZBGL;GRANT DBA TO ZBGL;ALTER USER ZBGL DEFAULT TABLESPACE WXGL_DATA1 TEMPORARY TABLESPACE TEMPORARY_DATA;

設定五個復原段的SQL語句如下：

SELECT SEGMENT_NAME FROM DBA_ROLLBACK_SEGS WHERE INITIAL_EXTENT < 512000 ANDUPPPER(OWNER) = 'PUBLIC';SELECT UPPER(STATUS) FROM DBA_ROLLBACK_SEGS WHERE UPPER(SEGMENT_NAME) = ''ALTER ROLLBACK SEGMENT RB1 OFFLINE;ALTER ROLLBACK SEGMENT RB2 OFFLINE;ALTER ROLLBACK SEGMENT RB3 OFFLINE;ALTER ROLLBACK SEGMENT RB4 OFFLINE;ALTER ROLLBACK SEGMENT RB5 OFFLINE;DROP ROLLBACK SEGMENT RB1;DROP ROLLBACK SEGMENT RB2;DROP ROLLBACK SEGMENT RB3;DROP ROLLBACK SEGMENT RB4;DROP ROLLBACK SEGMENT RB5;CREATE PUBLIC ROLLBACK SEGMENT RB1 TABLESPACE ROLLBACK_DATA STORAGE (INITIAL 512000 NEXT 512000 MAXEXTENTS 121);CREATE PUBLIC ROLLBACK SEGMENT RB2 TABLESPACE ROLLBACK_DATA STORAGE (INITIAL 512000 NEXT 512000 MAXEXTENTS 121);CREATE PUBLIC ROLLBACK SEGMENT RB3 TABLESPACE ROLLBACK_DATA STORAGE (INITIAL 512000 NEXT 512000 MAXEXTENTS 121);CREATE PUBLIC ROLLBACK SEGMENT RB4 TABLESPACE ROLLBACK_DATA STORAGE (INITIAL 512000 NEXT 512000 MAXEXTENTS 121);CREATE PUBLIC ROLLBACK SEGMENT RB5 TABLESPACE ROLLBACK_DATA STORAGE (INITIAL 512000 NEXT 512000 MAXEXTENTS 121);ALTER ROLLBACK SEGMENT RB1 ONLINE;ALTER ROLLBACK SEGMENT RB2 ONLINE;ALTER ROLLBACK SEGMENT RB3 ONLINE;ALTER ROLLBACK SEGMENT RB4 ONLINE;ALTER ROLLBACK SEGMENT RB5 ONLINE;COMMIT;

將資料量大的庫存表等放在一簇內的SQL語句如下：

KCB='CREATE TABLE QC_KCB( ' +' CKNM NUMBER(8) ,' +' QCNM NUMBER(10) ,' +' CKKC NUMBER(12,2),' +' SNCKKC NUMBER(12,2),' +' LDJ NUMBER(12,2),' +' BZ VARCHAR(100),' +' PRIMARY KEY(CKNM,QCNM))' +' TABLESPACE WXGL_DATA1 ' ;（大資料量的庫存表等放在WXGL_DATA1)QCFL = 'CREATE TABLE QC_QCFL ' + '(FLBH NUMBER(2) PRIMARY KEY,' + ' FLMC VARCHAR(20) ' + ' ) ' +' TABLESPACE WXGL_DATA2 ' ;(其他表放在WXGL_DATA2）

系統的基礎資料庫存表、名錄表大約有資料80M；一個單位一般每年收發300次，收發料單大約有資料50M；系統冗餘資料100M，系統輔助資料10M；因此，系統總共需要空間大約是240M，現在系統開闢資料空間500M，完全滿足儲存要求。由於系統使用了冗餘資料，在查詢常用資料時，避免了多表連接查詢的情況，這樣，雖然使用了更多的儲存空間，但查詢效率大幅度提高；同時，系統將需要經常連接查詢的資料放在一簇，即將存放這些資料的空間在物理上相鄰，這樣也使查詢速度大大提高。

另外在oracle7.33資料庫的Database目錄下有一個Initorcl.ora檔案，改變其中的設定也可以提高查詢統計速度。該檔案的內容如下：

## $Header: init.ora 1.2 94/10/18 16:12:36 gdudey Osd<desktop/netware> $ init.ora Copyr (c) 1991 Oracle################################################################################ Example INIT.ORA file## This file is provided by Oracle Corporation to help you customize# your RDBMS installation for your site. Important system parameters# are discussed, and example settings given.## Some parameter settings are generic to any size installation.# For parameters that require different values in different size# installations, three scenarios have been provided: SMALL, MEDIUM# and LARGE. Any parameter that needs to be tuned according to # installation size will have three settings, each one commented# according to installation size.## Use the following table to approximate the SGA size needed for the# three scenarious provided in this file:## -------Installation/Database Size------# SMALL MEDIUM LARGE# Block 2K 4500K 6800K 17000K# Size 4K 5500K 8800K 21000K## To set up a database that multiple instances will be using, place# all instance-specific parameters in one file, and then have all# of these files point to a master file using the IFILE command.# This way, when you change a public# parameter, it will automatically change on all instances. This is # necessary, since all instances must run with the same value for many# parameters. For example, if you choose to use private rollback segments, # these must be specified in different files, but since all gc_* # parameters must be the same on all instances, they should be in one file.## INSTRUCTIONS: Edit this file and the other INIT files it calls for# your site, either by using the values provided here or by providing# your own. Then place an IFILE= line into each instance-specific# INIT file that points at this file.###############################################################################db_name = oracledb_files = 20control_files = C:/ORAWIN95/DATABASE/ctl1orcl.oracompatible = 7.3.0.0.0db_file_multiblock_read_count = 8 # INITIAL# db_file_multiblock_read_count = 8 # SMALL # db_file_multiblock_read_count = 16 # MEDIUM # db_file_multiblock_read_count = 32 # LARGE db_block_buffers = 200 # INITIAL# db_block_buffers = 200 # SMALL # db_block_buffers = 550 # MEDIUM # db_block_buffers = 3200 # LARGE shared_pool_size = 3500000 # INITIAL# shared_pool_size = 3500000 # SMALL# shared_pool_size = 6000000 # MEDIUM# shared_pool_size = 9000000 # LARGElog_checkpoint_interval = 10000processes = 50 # INITIAL# processes = 50 # SMALL # processes = 100 # MEDIUM # processes = 200 # LARGE dml_locks = 100 # INITIAL# dml_locks = 100 # SMALL # dml_locks = 200 # MEDIUM # dml_locks = 500 # LARGElog_buffer = 8192 # INITIAL# log_buffer = 8192 # SMALL# log_buffer = 32768 # MEDIUM# log_buffer = 163840 # LARGEsequence_cache_entries = 10 # INITIAL# sequence_cache_entries = 10 # SMALL # sequence_cache_entries = 30 # MEDIUM # sequence_cache_entries = 100 # LARGE sequence_cache_hash_buckets = 10 # INITIAL# sequence_cache_hash_buckets = 10 # SMALL # sequence_cache_hash_buckets = 23 # MEDIUM # sequence_cache_hash_buckets = 89 # LARGE # audit_trail = true # if you want auditing# timed_statistics = true # if you want timed statisticsmax_dump_file_size = 10240 # limit trace file size to 5 Meg each # log_archive_start = true # if you want automatic archivinglog_archive_dest=%ORACLE_HOME%/DATABASE/ARCHIVE# define directories to store trace and alert filesbackground_dump_dest=%RDBMS73%/traceuser_dump_dest=%RDBMS73%/tracedb_block_size = 2048snapshot_refresh_processes = 1remote_login_passwordfile = shared

可以修改該檔案的db_file_multiblock_read_count，db_block_buffers，shared_pool_size，processes ，log_buffer，sequence_cache_entries，sequence_cache_hash_buckets等項（檔案中均有提示），根據需要和系統使用的資料庫伺服器的特點適當改大數值，可以提高查詢統計速度。這裡要注意的是，必須將 log_archive_start = true 項改為False，設定才會生效。

本文介紹的Oracle資料庫設定方法均在使用者實際使用中經過了嚴格測試，證明是有效和實用的。