9. Oracle深度学习笔记——性能调优汇总

欢迎转载，转载请标明出处：http://blog.csdn.net/notbaron/article/details/50621026

数据字典高速缓存命中率

数据字典高速缓存命中率如下：

select (sum(gets-getmisses-fixed))/sum(gets)"data dictionary hit ratio" from v$rowcache;

data dictionary hit ratio

-------------------------

.975728791

参考95~99%

1. 硬解析过程

alter session set sql_trace=true;

alter system flush shared_pool;

select table_name from user_tables whererownum<=5;

打开TRC文件如下：

PARSING IN CURSOR #140195096140104 len=50dep=0 uid=0 oct=3 lid=0 tim=700865302 hv=17921204 ad='97a9d708'sqlid='83dx9sc0j2x5n'

select table_name from user_tables whererownum<=5

END OF STMT

PARSE#140195096140104:c=106000,e=131031,p=2,cr=727,cu=0,mis=1,r=0,dep=0,og=1,plh=5472353,tim=700865301

….

再执行相同的命令如下，查看TRC文件如下：

*** 2016-01-30 17:55:48.379

CLOSE#140195096140104:c=0,e=100,dep=0,type=0,tim=1039278193

=====================

PARSING IN CURSOR #140195096140104 len=50dep=0 uid=0 oct=3 lid=0 tim=1039279623 hv=17921204 ad='97a9d708'sqlid='83dx9sc0j2x5n'

select table_name from user_tables whererownum<=5

END OF STMT

PARSE #140195096140104:c=1000,e=937,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=1,plh=5472353,tim=1039279622

……

发现命中。

使用tkprof处理一下：

具体使用，可以直接输入tkprof

打开结果文件如下：

select table_name

from

user_tables where rownum<=5

call count cpu elapsed disk query current rows

------- ------ -------- ---------- ---------- -------------------- ----------

Parse 2 0.05 0.06 0 5 0 0

Execute 2 0.00 0.00 0 0 0 0

Fetch 4 0.00 0.00 0 140 0 10

------- ------ -------- ---------- ---------- -------------------- ----------

total 8 0.06 0.07 0 145 0 10

Misses in library cacheduring parse: 1

2. 库高速缓存命中率

select sum(pinhits)/sum(pins) Library_cache_hit_ratiofrom v$librarycache;

LIBRARY_CACHE_HIT_RATIO

-----------------------

.927807473

参考大于99%

确定库高速缓存的效率

select namespace,pins,pinhits,reloads fromv$librarycache order by namespace;

NAMESPACE PINS PINHITS RELOADS

-------------------------------------------------------------------------- ---------- ----------

AUDIT POLICY 66 56 0

BODY 1301 1097 16

CLUSTER 1318 1298 0

CONTEXT POLICY 2 0 0

DBINSTANCE 0 0 0

DBLINK 0 0 0

DIRECTORY 51 41 0

EDITION 309 296 0

INDEX 4516 3082 72

OBJECT ID 0 0 0

PDB 0 0 0

QUEUE 10 6 0

RULESET 3 2 0

SCHEMA 0 0 0

SQL AREA 528694 508982 641

SQL AREA BUILD 0 0 0

SQL AREA STATS 6406 262 0

TABLE/PROCEDURE 57487 44691 3245

TEMPORARY INDEX 481 0 45

TEMPORARY TABLE 3361 0 1962

TRIGGER 528 405 0

USER PRIVILEGE 1 0 0

22 rows selected.

其中RELOAD表示SQL语句在老化后又被重新装入库。可以使用DBMS_SHARED_POOL程序强迫将SQL语句保留在共享池的库高速缓存组中。

V$LIBRARY_CACHE_MEMORY视图确定库高速缓存内存对象的数目

V$SHARED_POOL_ADVICE视图提供各种尺寸的共享池预期可节省的分析时间的信息

3. 优化库高速缓存

l CUSOR_SHARING

设置CURSOR_SHARING=FORCE来减轻非绑定变量的问题，强制进行绑定变量。

显示实例启动以来，硬分析数与执行数比较结果，如下：

select s.sid,s.value "Hard Parses", t.value"Executions Count" from v$sesstat s,v$sesstat t where s.sid=t.sid ands.statistic#=(select statistic# from v$statname where name='parse count(hard)') and t.statistic#=(select statistic# from v$statname where name='executecount') and s.value >0;

SID Hard Parses Executions Count

---------- ----------- ----------------

1 576 3895

9 15 77

15 11 82

16 8 51

21 325 4506

22 3 15

34 1 19

37 1 1

39 78 1680

45 12 116

47 1 2

67 6 80

12 rows selected.

l 参数CURSOR_SPACE_FOR_TIME已废弃

l SESSION_CACHED_CURSORS

一个session可以缓存多少个cursor，让后续相同的SQL语句不再打开游标,从而避免软解析的过程来提高性能。(绑定变量是解决硬解析的问题)，软解析同硬解析一样,比较消耗资源.所以这个参数非常重要。

session_cached_cursors这个参数是控制sessioncursor cache的大小的。session_cached_cursors定义了session cursor cache中存储的cursor的个数。这个值越大，则会消耗的内存越多。设置在OPEN_CURSORS和会话中使用的游标的数目之间。

查看游标命中次数

select name,value from v$sysstat where name like'%cursor%';

NAME VALUE

--------------------------------------------------------------------------

opened cursors cumulative 456915

opened cursors current 33

pinned cursors current 15

session cursor cache hits 151457

session cursor cache count 6346

cursor reload failures 14

cursor authentications 1219

7 rows selected.

查看总的分析次数：

select name,value from v$sysstat where name like'%parse%';

NAME VALUE

--------------------------------------------------------------------------

ADG parselock X get attempts 0

ADG parselock X get successes 0

parse time cpu 1571

parse time elapsed 4723

parse count (total) 108089

parse count (hard) 8458

parse count (failures) 377

parse count (describe) 21

8 rows selected.

4. 确定要定在共享池中的对象

select type,count(*) objects,sum(decode(kept,'YES',1,0))kept, sum(loads)-count(*) reloads from v$db_object_cache group by type order byobjects desc;

TYPE OBJECTS KEPT RELOADS

-------------------------------------------------------------------------- ---------- ----------

CURSOR 6419 0 499

CURSOR STATS 1909 1114 0

TABLE 691 30 884

PACKAGE 225 0 -120

TYPE 163 0 -48

INDEX 82 7 83

MULTI-VERSIONED OBJECT 64 62 3

PACKAGE BODY 43 0 3

Optimizer Finding 43 0 0

SEQUENCE 28 0 -18

SYNONYM 28 0 -21

LIBRARY 23 0 -19

TRIGGER 21 0 0

Optimizer Directive Owner 16 0 0

FUNCTION 13 0 3

NONE 13 0 -13

CLUSTER 7 6 0

TYPE BODY 4 0 0

PROCEDURE 4 0 -2

SCHEDULER CLASS 3 0 0

SCHEDULER JOB 3 0 0

SCHEDULER GLOBAL ATTRIBUTE 2 0 0

QUEUE 2 0 0

PUB SUB INTERNAL INFORMATION 2 0 1

OPERATOR 2 0 0

AUDIT POLICY 2 0 0

SCHEDULER PROGRAM 1 0 0

SCHEDULER WINDOW 1 0 0

RULE EVALUATION CONTEXT 1 0 -1

SCHEDULER GROUP 1 0 0

RESOURCE MANAGER CONSUMER GROUP 1 0 0

EDITION 1 0 0

RULESET 1 0 0

34 rows selected.

共享池中钉住程序包，命令如下：

EXECUTE SYS.DBMS_SHARED_POOL.KEEP（object_name,object_type）;

查看数据库中程序包的数量如下：

select count(*) from v$db_object_cache wheretype='PACKAGE';

COUNT(*)

----------

223

共享池中钉住的内存量：

select sum(sharable_mem) from v$db_object_cache wheretype='PACKAGE';

SUM(SHARABLE_MEM)

-----------------

2349720

5. 数据缓存优化相关

缓冲区高速缓存的命中率：

selectname,physical_reads,db_block_gets,consistent_gets,1-(physical_reads/(db_block_gets+consistent_gets))"HitRatio" from v$buffer_pool_statistics;

NAME PHYSICAL_READS DB_BLOCK_GETSCONSISTENT_GETS HitRatio

-------------------- --------------------------- --------------- ----------

DEFAULT 18216 159266 2316657 .992642744

可以使用v$db_cache_advice,v$sga_cache_advice 来优化缓冲区。

当然，事物都存在两极，如果存在大量全表扫描或数据仓库，命中率可能远低于100%，但并不一定是坏事。

使用多个池，保留缓冲池（keep）、回收缓冲池(recycle)、默认缓冲池(default)。

确定保留缓冲区高速缓存的备选

select obj object,count(1) buffers,avg(tch)average_touch_count from x$bh where lru_flag=8 group by obj havingavg(tch)>5 and count(1)>25;

将某个表放入到保留缓冲区中，命令如下：

alter table test1 storage (buffer_poolkeep);

6. PGA

查看数据库中PGA使用情况：

select s.value,s.sid,a.username from v$sesstats,v$statname n,v$session a where n.statistic#=s.statistic# and name='sessionpga memory' and s.sid=a.sid order by s.value;

VALUE SID USERNAME

---------- ----------------------------------------

766056 40

766056 17

766056 3

766056 4

766056 47

766056 18

766056 20

766056 6

766056 7

766056 28

766056 10

766056 24

774456 19

774456 8

774456 5

790912 2

962664 46

1421416 22

1486952 36

1535432 30

1618024 16

1618024 31

1749096 15

2093792 14

2330264 13

2330264 11

2338920 44

2609464 9

3395896 38

4911624 1 SYS

8827016 12

12636472 21

62238824 23

33 rows selected.

查看PGA状态:

select * from v$pgastat;

NAME VALUE UNIT CON_ID

-------------------------------------------------------------------------- ------------ ----------

aggregate PGA target parameter 377487360 bytes 0

aggregate PGA auto target 232538112 bytes 0

global memory bound 75497472 bytes 0

total PGA inuse 118634496 bytes 0

total PGA allocated 154334208 bytes 0

maximum PGA allocated 342854656 bytes 0

total freeable PGA memory 14745600 bytes 0

process count 42 0

max processes count 63 0

PGA memory freed back to OS 511049728 bytes 0

total PGA used for auto workareas 0 bytes 0

maximum PGA used for auto workareas 5123072 bytes 0

total PGA used for manual workareas 0 bytes 0

maximum PGA used for manual workareas 0 bytes 0

over allocation count 0 0

bytes processed 1479985152 bytes 0

extra bytes read/written 0 bytes 0

cache hit percentage 100percent 0

recompute count (total) 1436 0

19 rows selected.

查看PGA效能

select low_optimal_size/1024 "Low (K)",(high_optimal_size+1)/1024 "High (K)",optimal_executions"Optimal",onepass_executions "1-pass",multipasses_executions ">1 Pass" from v$sql_workarea_histogram wheretotal_executions <> 0;

Low (K) High (K) Optimal 1-pass >1 Pass

---------- ---------- ---------- --------------------

2 4 13821 0 0

64 128 124 0 0

128 256 35 0 0

256 512 14 0 0

512 1024 473 0 0

1024 2048 728 0 0

2048 4096 6 0 0

7 rows selected.

最佳方法来执行排序。

查看每个进程的PGA使用情况：

select pid,category,allocated,used fromv$process_memory;

PID CATEGORY ALLOCATED USED

---------- --------------- --------------------

2 PL/SQL 1976 224

2 Other 972532

3 PL/SQL 1976 224

3 Other 947740

4 PL/SQL 1976 224

4 Other 947740

5 PL/SQL 1976 224

5 Other 956108

6 PL/SQL 1976 224

6 Other 947740

7 SQL 43936 8440

7 PL/SQL 26128 21608

7 Freeable 917504 0

7 Other 3976580

8 PL/SQL 1976 224

8 Other 956108

9 SQL 0 0

9 PL/SQL 1976 224

9 Freeable 65536 0

9 Other 2791116

10PL/SQL 1976 224

10Other 947740

11PL/SQL 1976 224

11Freeable 131072 0

11Other 2511916

12PL/SQL 1976 224

12Freeable 786432 0

12Other 9008508

13PL/SQL 1976 224

13Freeable 393216 0

13Other 2511916

14PL/SQL 1976 224

14Freeable 393216 0

14Other 2275412

15SQL 2000 296

15PL/SQL 1976 224

15Freeable 262144 0

15Other 1928780

16SQL 0 0

16PL/SQL 1976 224

16Freeable 131072 0

16Other 1799708

17PL/SQL 1976 224

17Other 947740

18PL/SQL 1976 224

18Other 947740

19PL/SQL 1976 224

19Other 956108

20PL/SQL 1976 224

20Other 947740

21SQL 36576 0

21PL/SQL 312848 294744

21Freeable 3276800 0

21Other 12470676

22SQL 0 0

22PL/SQL 1976 224

22Other 1603100

23Other 1408468

24PL/SQL 1976 224

24Other 62420508

25Other 687572

26PL/SQL 1976 224

26Other 947740

27PL/SQL 1976 224

27Other 947740

28PL/SQL 1976 224

28Other 947740

29PL/SQL 1976 224

29Other 947740

30SQL 3024 1256

30PL/SQL 1976 224

30Other 1796684

31SQL 7096 800

31PL/SQL 1976 224

31Freeable 131072 0

31Other 1710020

33SQL 0 0

33PL/SQL 1976 224

33Freeable 131072 0

33Other 1668636

34SQL 11904 2728

34PL/SQL 20704 16816

34Freeable 7340032 0

34Other 3546916

35Other 687572

36Other 687572

37Other 687572

38Other 687572

39SQL 0 0

39PL/SQL 1976 224

39Other 947740

41SQL 3000 0

41PL/SQL 264056 257616

41Freeable 589824 0

41Other 2255524

42SQL 5760 3552

42PL/SQL 1976 224

42Other 1138588

98 rows selected.

评估实例启动以来，使用最佳、1遍、多遍PGA内存尺寸的工作区比例：

select name profile,cntcount,decode(total,0,0,round(cnt*100/total)) percentage from (select name,valuecnt,(sum(value) over()) total from v$sysstat where name like 'workarea exec%');

PROFILE COUNT PERCENTAGE

-------------------------------------------------------------------------- ----------

workarea executions - optimal 15231 100

workarea executions - onepass &nbs

再分享一下我老师大神的人工智能教程吧。零基础！通俗易懂！风趣幽默！还带黄段子！希望你也加入到我们人工智能的队伍中来！https://blog.csdn.net/jiangjunshow

9 Oracle深度学习笔记——性能调优汇总