PGA SGA db_cache 大小如何设置调优
v$sga_target_advice;
V$SGA_TARGET_ADVICE provides information about the SGA_TARGET initialization parameter.
| Column | Datatype | Description |
|---|---|---|
SGA_SIZE |
NUMBER |
Size of the SGA |
SGA_SIZE_FACTOR |
NUMBER |
Ratio between the SGA_SIZE and the current size of the SGA |
ESTD_DB_TIME |
NUMBER |
Estimated DB_TIME for this SGA_SIZE |
ESTD_DB_TIME_FACTOR |
NUMBER |
Ratio between ESTD_DB_TIME and DB_TIME for the current size of the SGA |
ESTD_PHYSICAL_READS |
NUMBER |
Estimated number of physical reads |
SQL> select * from v$sga_target_advice;
SGA_SIZE SGA_SIZE_FACTOR ESTD_DB_TIME ESTD_DB_TIME_FACTOR ESTD_PHYSICAL_READS
---------- --------------- ------------ ------------------- -------------------
240 .5 382 1.0241 23258
360 .75 373 1 16301
480 1 373 1 16213
600 1.25 373 1 16213
720 1.5 373 1 16213
840 1.75 373 1 16213
960 2 373 1 16213
解读:
1 原始的SGA值为：480M
2 当大于这个数值的时候我们发现无论是DB_TIME或是物理读都没有性能的提升!
3 当设置为0.75倍*480M也就是360M时,我们发现物理读数的性能有小小的性能降低
4 当设置为0.5倍*480M的时候,我们发现DB_TIME性能下降2.41% 而物理读数比原有的高出:23258-16213=7045
从上面的建议来看 我们设置的SGA480挺合理的!
SGA相关重要视图
select * from v$sgainfo;--各部门分配情况
select * from v$sgastat;--详细
select * from v$sga_dynamic_components;--动态
select t.component,
t.current_size / 1024 / 1024 || 'M' "当前大小(MB)",
MIN_SIZE / 1024 / 1024 || 'M' "MIN_SIZE(MB)",
MAX_SIZE / 1024 / 1024 || 'M' "MAX_SIZE(MB)"
from v$sga_dynamic_components t;
SQL> select t.component,
2 t.current_size / 1024 / 1024 || 'M' "当前大小(MB)",
3 MIN_SIZE / 1024 / 1024 || 'M' "MIN_SIZE(MB)",
4 MAX_SIZE / 1024 / 1024 || 'M' "MAX_SIZE(MB)"
5 from v$sga_dynamic_components t;
COMPONENT 当前大小(MB) MIN_SIZE(MB) MAX_SIZE(MB)
---------------------------------------------------------------- ----------------------------------------- ----------------------------------------- -----------------------------------------
shared pool 224M 224M 224M
large pool 4M 4M 4M
java pool 4M 4M 4M
streams pool 0M 0M 0M
DEFAULT buffer cache 240M 240M 240M
KEEP buffer cache 0M 0M 0M
RECYCLE buffer cache 0M 0M 0M
DEFAULT 2K buffer cache 0M 0M 0M
DEFAULT 4K buffer cache 0M 0M 0M
DEFAULT 8K buffer cache 0M 0M 0M
DEFAULT 16K buffer cache 0M 0M 0M
DEFAULT 32K buffer cache 0M 0M 0M
Shared IO Pool 0M 0M 0M
ASM Buffer Cache 0M 0M 0M
14 rows selected
SQL> --DEFAULT buffer cache 为240M
v$pga_target_advice
V$PGA_TARGET_ADVICE predicts how the cache hit percentage and over allocation count statistics displayed by the V$PGASTAT performance view would be impacted if the value of the PGA_AGGREGATE_TARGET parameter is changed. The prediction is performed for various values of the PGA_AGGREGATE_TARGET parameter, selected around its current value. The advice statistic is generated by simulating the past workload run by the instance.
The content of the view is empty if PGA_AGGREGATE_TARGET is not set. In addition, the content of this view is not updated if the STATISTICS_LEVEL parameter is set to BASIC. Base statistics for this view are reset at instance startup and when the value of the PGA_AGGREGATE_TARGET initialization parameter is dynamically modified.
| Column | Datatype | Description |
|---|---|---|
PGA_TARGET_FOR_ESTIMATE |
NUMBER |
Value of PGA_AGGREGATE_TARGET for this prediction (in bytes) |
PGA_TARGET_FACTOR |
NUMBER |
PGA_TARGET_FOR_ESTIMATE / the current value of the PGA_AGGREGATE_TARGET parameter |
ADVICE_STATUS |
VARCHAR2(3) |
Indicates whether the advice is enabled (ON) or disabled (OFF) depending on the value of the STATISTICS_LEVEL parameter |
BYTES_PROCESSED |
NUMBER |
Total bytes processed by all the work areas considered by this advice (in bytes) |
ESTD_EXTRA_BYTES_RW |
NUMBER |
Estimated number of extra bytes which would be read or written if PGA_AGGREGATE_TARGETwas set to the value of the PGA_TARGET_FOR_ESTIMATE column. This number is derived from the estimated number and size of work areas which would run in one-pass (or multi-pass) for that value of PGA_AGGREGATE_TARGET. |
ESTD_PGA_CACHE_HIT_PERCENTAGE |
NUMBER |
Estimated value of the cache hit percentage statistic when PGA_AGGREGATE_TARGET equalsPGA_TARGET_FOR_ESTIMATE. This column is derived from the above two columns and is equal to BYTES_PROCESSED / (BYTES_PROCESSED + ESTD_EXTRA_BYTES_RW) |
ESTD_OVERALLOC_COUNT |
NUMBER |
Estimated number of PGA memory over-allocations if the value of PGA_AGGREGATE_TARGET is set to PGA_TARGET_FOR_ESTIMATE. A nonzero value means that PGA_TARGET_FOR_ESTIMATE is not large enough to run the work area workload. Hence, the DBA should not setPGA_AGGREGATE_TARGET to PGA_TARGET_FOR_ESTIMATE since Oracle will not be able to honor that target. |
SQL> Select pga_target_for_estimate / 1024 / 1024 "PGA(MB)",
2 pga_target_factor "与当前PGA比(倍)",
3 estd_pga_cache_hit_percentage 命中率,
4 estd_overalloc_count 超载
5 FROM v$pga_target_advice order by 1 desc;
PGA(MB) 与当前PGA比(倍) 命中率 超载
---------- -------------------- ---------- ----------
2560 8 100 0
1920 6 100 0
1280 4 100 0
960 3 100 0
640 2 100 0
576 1.8 100 0
512 1.6 100 0
448 1.4 100 0
384 1.2 100 0
320 1 100 0
240 0.75 100 0
160 0.5 100 0
80 0.25 80 5
40 0.125 80 5
14 rows selected
SQL>
解读:
1 我们原有的PGA为320M
2 当PGA降低到50％ 160M我们的命中率依然100％
从上面的建议值我们可以降低我们的PGA，缓解内存紧张的压力~!~
v$db_cache_advice
下面是关于这个视图的结构描述：
| 字段 |
数据类型 |
描述 |
| ID |
NUMBER |
缓冲池标识号（从1到8，1-6对应于DB_nK_CACHE_SIZE，DB_CACHE_SIZE与系统标准块尺寸的序号相关，如DB_BLOCK_SIZE为8K，则DB_CACHE_SIZE的标识号为3（2,4,8…）。7是DB_KEEP_CACHE_SIZE，8是DB_RECYCLE_CACHE_SIZE） |
| NAME |
VARCHAR2(20) |
缓冲池名称 |
| BLOCK_SIZE |
NUMBER |
缓冲池块尺寸（字节为单位） |
| ADVICE_STATUS |
VARCHAR2(3) |
建议器状态：ON表示建议器在运行；OFF表示建议器已经关闭。当建议器关闭了，视图中的数据是上一次打开所统计得出的。 |
| SIZE_FOR_ESTIMATE |
NUMBER |
预测性能数据的Cache大小（M为单位） |
| SIZE_FACTOR |
NUMBER |
预测的Cache大小因子（即与当前大小的比例） |
| BUFFERS_FOR_ESTIMATE |
NUMBER |
预测性能数据的Cache大小（缓冲块数） |
| ESTD_PHYSICAL_READ_FACTOR |
NUMBER |
这一缓冲大小时，物理读因子，它是如果缓冲大小为SIZE_FOR_ESTIMATE时，建议器预测物理读数与当前实际物理读数的比率值。如果当前物理读数为0，这个值为空。 |
| ESTD_PHYSICAL_READS |
NUMBER |
如果缓冲大小为SIZE_FOR_ESTIMATE时，建议器预测物理读数。 |
SQL:
select size_for_estimate,
size_factor,
estd_physical_read_factor,
estd_physical_reads
from v$db_cache_advice
where name = 'DEFAULT';
SQL> select size_for_estimate,
2 size_factor,
3 estd_physical_read_factor,
4 estd_physical_reads
5 from v$db_cache_advice
6 where name = 'DEFAULT';
SIZE_FOR_ESTIMATE SIZE_FACTOR ESTD_PHYSICAL_READ_FACTOR ESTD_PHYSICAL_READS
----------------- ----------- ------------------------- -------------------
24 0.1 1.4371 23611
48 0.2 1.1667 19168
72 0.3 1.0762 17683
96 0.4 1.07 17581
120 0.5 1.0062 16532
144 0.6 1 16430
168 0.7 1 16430
192 0.8 1 16430
216 0.9 1 16430
240 1 1 16430
264 1.1 1 16430
288 1.2 1 16430
312 1.3 1 16430
336 1.4 1 16430
360 1.5 1 16430
384 1.6 1 16430
408 1.7 1 16430
432 1.8 1 16430
456 1.9 1 16430
480 2 1 16430
20 rows selected
SQL>
解读:
1 我们当前的db_cache 为:240M
2 当值为0.6倍*240M=144M时我们发现物理读数的性能无差
如果手动管理内存的话可以调整为144M !当然内存资源不要这么扣,有就用......
命中率怎么查?
select 1-(sum(decode(name, 'physical reads', value, 0))/
(sum(decode(name, 'db block gets', value, 0))+
(sum(decode(name, 'consistent gets', value, 0))))) "Buffer Hit Ratio"
from v$sysstat;