« 关于shared pool的深入探讨(二) |Blog首页| 使用SQL_TRACE进行数据库诊断 »
关于shared pool的深入探讨(一)
作者： eygle | English 【 版权声明：转载时请务必以超链接形式标明文章 原始出处和作者信息及 本声明】
链接： http://www.eygle.com/archives/2004/10/shared_pool-1.html
关于shared pool的设置一直是一个争议较多的内容.
很多文章上说,shared pool设置过大会带来额外的管理上的负担，从而在某些条件下会导致性能的下降.
那么这个管理上的负担指的是什么内容呢?
本文对这个内容作一定的深入探讨.
本文只涉及一个方面,后续的文章将从其他方面继续讨论.
基础知识:
我们可以通过如下命令转储shared pool共享内存的内容:
|
SQL> alter session set events 'immediate trace name heapdump level 2'; Session altered. 本测试中引用的两个trace文件: 9i: SQL> @gettrcname TRACE_FILE_NAME
--------------------------------------------------------------------------------
/opt/oracle/admin/hsjf/udump/hsjf_ora_24983.trc 8i: SQL> @gettrcname TRACE_FILE_NAME
--------------------------------------------------------------------------------
/usr/oracle8/admin/guess/udump/guess_ora_22038.trc
|
Shared Pool通过free list管理free块,Free List按不同size划分Bucket
在Oracle8i中,不同bucket的size范围如下所示(size显示的是下边界):
| oracle:/usr/oracle8/admin/guess/udump>cat guess_ora_22038.trc|grep Bucket
Bucket 0 size=44
Bucket 1 size=76
Bucket 2 size=140
Bucket 3 size=268
Bucket 4 size=524
Bucket 5 size=1036
Bucket 6 size=2060
Bucket 7 size=4108
Bucket 8 size=8204
Bucket 9 size=16396
Bucket 10 size=32780
|
我们注意,在这里，小于76的块都位于Bucket 0上;大于32780的块,都在Bucket 10上
初始的,数据库启动以后,shared pool多数是连续内存块
当空间分配使用以后,内存块开始被分割,碎片开始出现，Bucket列表开始变长
Oracle请求shared pool空间时，首先进入相应的Bucket进行查找。
如果找不到,则转向下一个非空的bucket,获取第一个chunk。
分割这个chunk,剩余部分会进入相应的Bucket,进一步增加碎片。
最终的结果是,Bucket 0上的内存块会越来越多,越来越碎小
(在我这个测试的小型的数据库上,Bucket 0上的碎片已经达到9030个
而shared_pool_size设置仅为150M)
通常如果每个Bucket上的chunk多余2000个，就被认为是share pool碎片过多
Shared Pool的碎片过多，是Shared pool产生性能问题的主要原因。
碎片过多会导致Search Free List的时间过长，从而使shared pool latch被长时间持有，导致更多的Latch竞争。
而在大多数情况下,我们请求的都是相对小的chunk,这样搜索Bucket 0往往消耗了大量的时间以及资源
这可能导致share pool Latch被长时间的持有,导致更多的share pool竞争
所以在Oracle9i之前,如果盲目的增大shared_pool_size或设置过大的shared_pool_size，往往会适得其反
我们看一下Oracle9i中的处理方式:
|
[oracle@jumper oracle]$ sqlplus "/ as sysdba" SQL*Plus: Release 9.2.0.3.0 - Production on Wed Aug 18 22:13:07 2004 Copyright (c) 1982, 2002, Oracle Corporation. All rights reserved.
Connected to:
Oracle9i Enterprise Edition Release 9.2.0.3.0 - Production
With the Partitioning, OLAP and Oracle Data Mining options
JServer Release 9.2.0.3.0 - Production SQL> alter session set events 'immediate trace name heapdump level 2'; Session altered. SQL> @gettrcname TRACE_FILE_NAME
--------------------------------------------------------------------------------
/opt/oracle/admin/hsjf/udump/hsjf_ora_24983.trc SQL>
SQL> !
[oracle@jumper oracle]$ cd $admin
[oracle@jumper udump]$ cat hsjf_ora_24983.trc|grep Bucket
Bucket 0 size=16
Bucket 1 size=20
Bucket 2 size=24
Bucket 3 size=28
Bucket 4 size=32
Bucket 5 size=36
Bucket 6 size=40
Bucket 7 size=44
Bucket 8 size=48
Bucket 9 size=52
Bucket 10 size=56
Bucket 11 size=60
Bucket 12 size=64
Bucket 13 size=68
Bucket 14 size=72
Bucket 15 size=76
Bucket 16 size=80
Bucket 17 size=84
Bucket 18 size=88
Bucket 19 size=92
Bucket 20 size=96
Bucket 21 size=100
Bucket 22 size=104
Bucket 23 size=108
Bucket 24 size=112
Bucket 25 size=116
Bucket 26 size=120
Bucket 27 size=124
Bucket 28 size=128
Bucket 29 size=132
Bucket 30 size=136
Bucket 31 size=140
Bucket 32 size=144
Bucket 33 size=148
Bucket 34 size=152
Bucket 35 size=156
Bucket 36 size=160
Bucket 37 size=164
Bucket 38 size=168
Bucket 39 size=172
Bucket 40 size=176
Bucket 41 size=180
Bucket 42 size=184
Bucket 43 size=188
Bucket 44 size=192
Bucket 45 size=196
Bucket 46 size=200
Bucket 47 size=204
Bucket 48 size=208
Bucket 49 size=212
Bucket 50 size=216
Bucket 51 size=220
Bucket 52 size=224
Bucket 53 size=228
Bucket 54 size=232
Bucket 55 size=236
Bucket 56 size=240
Bucket 57 size=244
Bucket 58 size=248
Bucket 59 size=252
Bucket 60 size=256
Bucket 61 size=260
Bucket 62 size=264
Bucket 63 size=268
Bucket 64 size=272
Bucket 65 size=276
Bucket 66 size=280
Bucket 67 size=284
Bucket 68 size=288
Bucket 69 size=292
Bucket 70 size=296
Bucket 71 size=300
Bucket 72 size=304
Bucket 73 size=308
Bucket 74 size=312
Bucket 75 size=316
Bucket 76 size=320
Bucket 77 size=324
Bucket 78 size=328
Bucket 79 size=332
Bucket 80 size=336
Bucket 81 size=340
Bucket 82 size=344
Bucket 83 size=348
Bucket 84 size=352
Bucket 85 size=356
Bucket 86 size=360
Bucket 87 size=364
Bucket 88 size=368
Bucket 89 size=372
Bucket 90 size=376
Bucket 91 size=380
Bucket 92 size=384
Bucket 93 size=388
Bucket 94 size=392
Bucket 95 size=396
Bucket 96 size=400
Bucket 97 size=404
Bucket 98 size=408
Bucket 99 size=412
Bucket 100 size=416
Bucket 101 size=420
Bucket 102 size=424
Bucket 103 size=428
Bucket 104 size=432
Bucket 105 size=436
Bucket 106 size=440
Bucket 107 size=444
Bucket 108 size=448
Bucket 109 size=452
Bucket 110 size=456
Bucket 111 size=460
Bucket 112 size=464
Bucket 113 size=468
Bucket 114 size=472
Bucket 115 size=476
Bucket 116 size=480
Bucket 117 size=484
Bucket 118 size=488
Bucket 119 size=492
Bucket 120 size=496
Bucket 121 size=500
Bucket 122 size=504
Bucket 123 size=508
Bucket 124 size=512
Bucket 125 size=516
Bucket 126 size=520
Bucket 127 size=524
Bucket 128 size=528
Bucket 129 size=532
Bucket 130 size=536
Bucket 131 size=540
Bucket 132 size=544
Bucket 133 size=548
Bucket 134 size=552
Bucket 135 size=556
Bucket 136 size=560
Bucket 137 size=564
Bucket 138 size=568
Bucket 139 size=572
Bucket 140 size=576
Bucket 141 size=580
Bucket 142 size=584
Bucket 143 size=588
Bucket 144 size=592
Bucket 145 size=596
Bucket 146 size=600
Bucket 147 size=604
Bucket 148 size=608
Bucket 149 size=612
Bucket 150 size=616
Bucket 151 size=620
Bucket 152 size=624
Bucket 153 size=628
Bucket 154 size=632
Bucket 155 size=636
Bucket 156 size=640
Bucket 157 size=644
Bucket 158 size=648
Bucket 159 size=652
Bucket 160 size=656
Bucket 161 size=660
Bucket 162 size=664
Bucket 163 size=668
Bucket 164 size=672
Bucket 165 size=676
Bucket 166 size=680
Bucket 167 size=684
Bucket 168 size=688
Bucket 169 size=692
Bucket 170 size=696
Bucket 171 size=700
Bucket 172 size=704
Bucket 173 size=708
Bucket 174 size=712
Bucket 175 size=716
Bucket 176 size=720
Bucket 177 size=724
Bucket 178 size=728
Bucket 179 size=732
Bucket 180 size=736
Bucket 181 size=740
Bucket 182 size=744
Bucket 183 size=748
Bucket 184 size=752
Bucket 185 size=756
Bucket 186 size=760
Bucket 187 size=764
Bucket 188 size=768
Bucket 189 size=772
Bucket 190 size=776
Bucket 191 size=780
Bucket 192 size=784
Bucket 193 size=788
Bucket 194 size=792
Bucket 195 size=796
Bucket 196 size=800
Bucket 197 size=804
Bucket 198 size=808
Bucket 199 size=812
Bucket 200 size=876
Bucket 201 size=940
Bucket 202 size=1004
Bucket 203 size=1068
Bucket 204 size=1132
Bucket 205 size=1196
Bucket 206 size=1260
Bucket 207 size=1324
Bucket 208 size=1388
Bucket 209 size=1452
Bucket 210 size=1516
Bucket 211 size=1580
Bucket 212 size=1644
Bucket 213 size=1708
Bucket 214 size=1772
Bucket 215 size=1836
Bucket 216 size=1900
Bucket 217 size=1964
Bucket 218 size=2028
Bucket 219 size=2092
Bucket 220 size=2156
Bucket 221 size=2220
Bucket 222 size=2284
Bucket 223 size=2348
Bucket 224 size=2412
Bucket 225 size=2476
Bucket 226 size=2540
Bucket 227 size=2604
Bucket 228 size=2668
Bucket 229 size=2732
Bucket 230 size=2796
Bucket 231 size=2860
Bucket 232 size=2924
Bucket 233 size=2988
Bucket 234 size=3052
Bucket 235 size=3116
Bucket 236 size=3180
Bucket 237 size=3244
Bucket 238 size=3308
Bucket 239 size=3372
Bucket 240 size=3436
Bucket 241 size=3500
Bucket 242 size=3564
Bucket 243 size=3628
Bucket 244 size=3692
Bucket 245 size=3756
Bucket 246 size=3820
Bucket 247 size=3884
Bucket 248 size=3948
Bucket 249 size=4012
Bucket 250 size=4108
Bucket 251 size=8204
Bucket 252 size=16396
Bucket 253 size=32780
Bucket 254 size=65548
|
我们看到,在Oracle9i中,Free Lists被划分为0~254,共255个Bucket
每个Bucket容纳的size范围
Bucket 0~199 容纳size以 4 递增
Bucket 200~249 容纳size以 64 递增
从Bucket 249开始,Oracle各Bucket步长进一步增加:
Bucket 249: 4012 ~4107 = 96
Bucket 250: 4108 ~8203 = 4096
Bucket 251: 8204 ~16395 = 8192
Bucket 252: 16396~32779 = 16384
Bucket 253: 32780~65547 = 32768
Bucket 254: >=65548
在Oracle9i中,对于小的chunk,Oracle增加了更多的Bucket来管理
0~199共200个Bucket,size以4为步长递增;200~249共50个Bucket,size以64递增.
这样每个Bucket中容纳的chunk数量大大减少,查找的效率得以提高.
这就是Oracle9i中shared pool管理的增强,通过这个算法的改进
Oracle8i中,过大shared pool带来的栓锁争用等性能问题在某种程度上得以解决.
下一页>>
-----
这篇 【 关于shared pool的深入探讨(一)】来自 www.eygle.com | 本站
版权| CSDN技术网摘| del.icio.us|
365Key
By eygle on 2004-10-31 09:29 | Comments (0) | Posted to Internal
站内相关主题:
关于shared pool的深入探讨(二)(0) October 31, 2004
关于shared pool的深入探讨(三)(0) October 31, 2004
关于shared pool的深入探讨(四)(0) October 31, 2004
关于shared pool的深入探讨(五)(2) October 31, 2004
关于shared pool的深入探讨(六)-高Latch竞争案例(0) October 31, 2004
