数据库的性能问题有哪些
本篇内容介绍了“数据库的性能问题有哪些”的有关知识,在实际案例的操作过程中,不少人都会遇到这样的困境,接下来就让小编带领大家学习一下如何处理这些情况吧!希望大家仔细阅读,能够学有所成!
创新互联公司坚持“要么做到,要么别承诺”的工作理念,服务领域包括:成都网站制作、成都网站建设、企业官网、英文网站、手机端网站、网站推广等服务,满足客户于互联网时代的河曲网站设计、移动媒体设计的需求,帮助企业找到有效的互联网解决方案。努力成为您成熟可靠的网络建设合作伙伴!
谓词越界常见发生在 where 谓词是时间字段的情况,总的来说统计信息记录的是一个过旧的时间,而 SQL 传入的时间是一个最新的时间范围(往往是
谓词越界就是 select 的谓词的条件不在统计信息 low_value 和 high_value 之间,在实际选择结果集要大于 CBO 记录的结果集数量,即实际的 selectivity 偏大,这种情况下 CBO 评估出来的 selectivity 会出现严重的偏差,导致 CBO 选错执行计划。
测试验证
下面做一组测试,从执行计划 cost 看谓词越界的发生过程,先插入部分数据:
DECLARE i INT; BEGIN i := 78179; WHILE(i < 100000) LOOP i := i + 1; INSERT INTO test_obj(object_id) VALUES(i); COMMIT; END LOOP; END; /
查看此时的 num_rows:
TEST@PROD1> select count(*) from test_obj; COUNT(*) ---------- 94283 TEST@PROD1> select max(object_ID),dump(max(object_id),16) from test_obj; MAX(OBJECT_ID) DUMP(MAX(OBJECT_ID),16) -------------- ---------------------------------------- 100000 Typ=2 Len=2: c3,b TEST@PROD1> select min(object_ID),dump(min(object_id),16) from test_obj; MIN(OBJECT_ID ) DUMP(MIN(OBJECT_ID),16) ------------------------------ ---------------------------------------- 2 Typ=2 Len=2: c1,3 --C103
不收集统计信息,此时统计列统计信息过旧,HIGH_VALUE 依然是原来的值 78179。
TEST@PROD1> select low_value ,high_value,num_distinct,num_nulls from DBA_TAB_COL_STATISTICS where table_name='TEST_OBJ' and owner='TEST'; Distinct Number LOW_VALUE HIGH_VALUE Values Nulls ------------------------------ ------------------------------ ------------ ---------- C103 C3085250 72,462(原值) 0
查询结果返回 2081 行结果集。
TEST@PROD1> select count(*) from test_obj where object_id between 78200 and 81000; COUNT(*) ---------- 2801 计算公式为: selectivity=((VAL2 - VAL1) / (HIGH_VALUE - LOW_VALUE)+2 / NUM_DISTINCT) * null_adjust null_adjust=(NUM_ROES - NUM_NULLS) / NUM_ROES 计算结果为: TEST@PROD1> select round(((81000-78200)/(100000-2)+2/94283)*(94283-0)/94283*94283) from dual; ROUND(((81000-78200)/(100000-2)+2/94283)*(94283-0)/94283*94283) --------------------------------------------------------------- 2642
查看结果集发现 dictionary 值为 1,这明显是一个错误的执行计划,由于统计信息过旧,已经低于谓词条件区间(谓词过界)导致 CBO 低估了查询成本。
TEST@PROD1> select count(*) from test_obj where object_id between 78200 and 81000; Execution Plan ---------------------------------------------------------- Plan hash value: 2217143630 ------------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | ------------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 1 | 5 | 289 (1)| 00:00:04 | | 1 | SORT AGGREGATE | | 1 | 5 | | | |* 2 | TABLE ACCESS FULL| TEST_OBJ | 1 | 5 | 289 (1)| 00:00:04 | ------------------------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------- 2 - filter("OBJECT_ID">=78200 AND "OBJECT_ID"<=81000) Statistics ---------------------------------------------------------- 1 recursive calls 0 db block gets 1117 consistent gets 0 physical reads 0 redo size 423 bytes sent via SQL*Net to client 419 bytes received via SQL*Net from client 2 SQL*Net roundtrips to/from client 0 sorts (memory) 0 sorts (disk) 1 rows processed
重新收集统计信息再次查看执行计划。
TEST@PROD1> exec dbms_stats.gather_table_stats('test','test_obj'); TEST@PROD1> select low_value ,high_value,num_distinct,num_nulls from DBA_TAB_COL_STATISTICS where table_name='TEST_OBJ' and owner='TEST'; Distinct Number LOW_VALUE HIGH_VALUE Values Nulls -------------------- -------------------- ------------ ---------- C103 C30B 94,283 0
此时统计信息 HIGH_VALUE 已经和最初计算的值相等,Typ=2 Len=2: c3,b。再次查看执行计划,此时 CBO 已经能够产生了正确的执行计划了。
执行计划为:
TEST@PROD1> select count(*) from test_obj where object_id between 78200 and 81000; Execution Plan ---------------------------------------------------------- Plan hash value: 2217143630 ------------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | ------------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 1 | 5 | 314 (1)| 00:00:04 | | 1 | SORT AGGREGATE | | 1 | 5 | | | |* 2 | TABLE ACCESS FULL| TEST_OBJ | 2642 | 13210 | 314 (1)| 00:00:04 | ------------------------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------- 2 - filter("OBJECT_ID">=78200 AND "OBJECT_ID"<=81000) Statistics ---------------------------------------------------------- 0 recursive calls 0 db block gets 1117 consistent gets 0 physical reads 0 redo size 423 bytes sent via SQL*Net to client 419 bytes received via SQL*Net from client 2 SQL*Net roundtrips to/from client 0 sorts (memory) 0 sorts (disk) 1 rows processed
谓词越界主要发生在大表,按照 Oracle 统计信息收集机制,表的数据变化量达到 10% 以上才会进行统计信息收集,大表不常收集统计信息就容易爆发谓词越界。
预防方式
可对关键表实行按谓词查询条件分区,即按天或者按月分区可规避此问题发生。
“数据库的性能问题有哪些”的内容就介绍到这里了,感谢大家的阅读。如果想了解更多行业相关的知识可以关注创新互联网站,小编将为大家输出更多高质量的实用文章!
网页标题:数据库的性能问题有哪些
文章来源:http://pwwzsj.com/article/pesdhs.html