首页 > 代码库 > oracle 优化 —— 分区表

oracle 优化 —— 分区表

一、分区表简介

  分区表类型:【范围分区】、【列表分区】 【hash分区】    【这些分区的组合分区】

    范围分区:以某一个范围进行分区。eg:时间段划分。

    列表分区:以某一些几个值进行分区。eg:地区分区,省份进行划分。

    hash分区:以hash算法进行分块。可以有效的消除io的竞争。 更多用在组合分区的子分区中。

    组合分区:11g前仅有两种组合分区  (range- *)      eg: 范围 -列表(月份地区),范围- hash 两种组合

         11g后新增四种。(range-range,list-list,list-hash,list-range) 考虑到兼容性等问题尽量使用 范围开头的组合分区。

  

  使用分区表优点:

    减少访问路径,提升性能外

     更方便的批量操作数据从而维护方便。

    不同的分区映射到磁盘以平衡I/O,改善整个系统性能。

 

二、分区表实战

  范围分区示例

  
 1 -- 范围分区示例 2 drop table range_part_tab purge; 3 --注意,此分区为范围分区 4  5 --例子1 6 create table range_part_tab (id number,deal_date date,area_code number,nbr number,contents varchar2(4000)) 7            partition by range (deal_date) 8            ( 9            partition p_201301 values less than (TO_DATE(2013-02-01, YYYY-MM-DD)),10            partition p_201302 values less than (TO_DATE(2013-03-01, YYYY-MM-DD)),11            partition p_201303 values less than (TO_DATE(2013-04-01, YYYY-MM-DD)),12            partition p_201304 values less than (TO_DATE(2013-05-01, YYYY-MM-DD)),13            partition p_201305 values less than (TO_DATE(2013-06-01, YYYY-MM-DD)),14            partition p_201306 values less than (TO_DATE(2013-07-01, YYYY-MM-DD)),15            partition p_201307 values less than (TO_DATE(2013-08-01, YYYY-MM-DD)),16            partition p_201308 values less than (TO_DATE(2013-09-01, YYYY-MM-DD)),17            partition p_201309 values less than (TO_DATE(2013-10-01, YYYY-MM-DD)),18            partition p_201310 values less than (TO_DATE(2013-11-01, YYYY-MM-DD)),19            partition p_201311 values less than (TO_DATE(2013-12-01, YYYY-MM-DD)),20            partition p_201312 values less than (TO_DATE(2014-01-01, YYYY-MM-DD)),21            partition p_201401 values less than (TO_DATE(2014-02-01, YYYY-MM-DD)),22            partition p_201402 values less than (TO_DATE(2014-03-01, YYYY-MM-DD)),23            partition p_max values less than (maxvalue)24            )25            ;26 27 28 --以下是插入2013年一整年日期随机数和表示福建地区号含义(591到599)的随机数记录,共有10万条,如下:29 insert into range_part_tab (id,deal_date,area_code,nbr,contents)30       select rownum,31              to_date( to_char(sysdate-365,J)+TRUNC(DBMS_RANDOM.VALUE(0,365)),J),32              ceil(dbms_random.value(591,599)),33              ceil(dbms_random.value(18900000001,18999999999)),34              rpad(*,400,*)35         from dual36       connect by rownum <= 100000;37 commit;38 39 40 41 --以下是插入2014年一整年日期随机数和表示福建地区号含义(591到599)的随机数记录,共有10万条,如下:42 insert into range_part_tab (id,deal_date,area_code,nbr,contents)43       select rownum,44              to_date( to_char(sysdate,J)+TRUNC(DBMS_RANDOM.VALUE(0,365)),J),45              ceil(dbms_random.value(591,599)),46              ceil(dbms_random.value(18900000001,18999999999)),47              rpad(*,400,*)48         from dual49       connect by rownum <= 100000;50 commit;51 52 53 ---添加一个全局索引、一个局部索引后,后面会提到分区操作对索引的影响。54 create index idx_part_id on range_part_tab (id) ;55 create index idx_part_nbr on range_part_tab (nbr) local;56 57 --统计信息系统一般会自动收集,这只是首次建成表后需要操作一下,以方便测试58 exec dbms_stats.gather_table_stats(ownname => LJB,tabname => RANGE_PART_TAB,estimate_percent => 10,method_opt=> for all indexed columns,cascade=>TRUE) ;  59 60 61 select min(deal_date),max(deal_date) from range_part_tab;62 63 --查看每个分区一共保存了多少条数据64 select count(*) from range_part_tab partition (p_201301);65 select count(*) from range_part_tab partition (p_201302);66 select count(*) from range_part_tab partition (p_201303);67 select count(*) from range_part_tab partition (p_201304);68 select count(*) from range_part_tab partition (p_201305);69 select count(*) from range_part_tab partition (p_201306);70 select count(*) from range_part_tab partition (p_201307);71 select count(*) from range_part_tab partition (p_201308);72 select count(*) from range_part_tab partition (p_201309);73 select count(*) from range_part_tab partition (p_201310);74 select count(*) from range_part_tab partition (p_201311);75 select count(*) from range_part_tab partition (p_201312);76 select count(*) from range_part_tab partition (p_max);
范围分区sql代码实战,脚本可以直接执行

  列表分区示例

  
 1 --列表分区示例 2 drop table list_part_tab purge; 3 --注意,此分区为列表分区 4 create table list_part_tab (id number,deal_date date,area_code number,nbr number,contents varchar2(4000)) 5            partition by list (area_code) 6            ( 7            partition p_591 values  (591), 8            partition p_592 values  (592), 9            partition p_593 values  (593),10            partition p_594 values  (594),11            partition p_595 values  (595),12            partition p_596 values  (596),13            partition p_597 values  (597),14            partition p_598 values  (598),15            partition p_599 values  (599),16            partition p_other values  (DEFAULT)17            )18            ;19            20 --以下是插入2013年一整年日期随机数和表示福建地区号含义(591到599)的随机数记录,共有10万条,如下:21 insert into list_part_tab (id,deal_date,area_code,nbr,contents)22       select rownum,23              to_date( to_char(sysdate-365,J)+TRUNC(DBMS_RANDOM.VALUE(0,365)),J),24              ceil(dbms_random.value(590,599)),25              ceil(dbms_random.value(18900000001,18999999999)),26              rpad(*,400,*)27         from dual28       connect by rownum <= 100000;29 commit;30 31 32 select count(*) from list_part_tab partition(p_591);33 select count(*) from list_part_tab partition(p_592);34 select count(*) from list_part_tab partition(p_593);35 select count(*) from list_part_tab partition(p_594);36 select count(*) from list_part_tab partition(p_595);37 select count(*) from list_part_tab partition(p_596);38 select count(*) from list_part_tab partition(p_597);39 select count(*) from list_part_tab partition(p_598);40 select count(*) from list_part_tab partition(p_599);41 select count(*) from list_part_tab partition(p_other);42 43 44 create index idx_list_part_id  on list_part_tab (id) ;45 create index idx_list_part_nbr on list_part_tab (nbr) local;46 47 --统计信息系统一般会自动收集,这只是首次建成表后需要操作一下,以方便测试48 exec dbms_stats.gather_table_stats(ownname => LJB,tabname => LIST_PART_TAB,estimate_percent => 10,method_opt=> for all indexed columns,cascade=>TRUE) ;  
列表分区sql脚本

  hash分区示例

 1 --散列分区示例 2 drop table hash_part_tab purge; 3 --注意,此分区HASH分区 4 create table hash_part_tab (id number,deal_date date,area_code number,nbr number,contents varchar2(4000)) 5             partition by hash (deal_date) 6             PARTITIONS 12 7             ; 8 --以下是插入2013年一整年日期随机数和表示福建地区号含义(591到599)的随机数记录,共有10万条,如下: 9 insert into hash_part_tab(id,deal_date,area_code,nbr,contents)10       select rownum,11              to_date( to_char(sysdate-365,J)+TRUNC(DBMS_RANDOM.VALUE(0,365)),J),12              ceil(dbms_random.value(590,599)),13              ceil(dbms_random.value(18900000001,18999999999)),14              rpad(*,400,*)15         from dual16       connect by rownum <= 100000;17 commit;18 19 20 21 --以下分区名是通过数据字典user_segments的partition_name查出来的,详见后面说明。22 ---每个分区存放多少数据23 select partition_name, 24        segment_type, 25        bytes,26        select count(*) from hash_part_tab partition(||partition_name||);27   from user_segments28  where segment_name =HASH_PART_TAB; 
hash分区sql脚本实战

  联合字段分区(两种联合起来进行分区)

  1 -- 范围分区示例  2 drop table range_part_mult_col_tab purge;  3 --注意,此分区为联合字段的范围分区  4   5 create table range_part_mult_col_tab (id number,deal_date date,area_code number,nbr number,contents varchar2(4000))  6            partition by range (area_code,deal_date)  7            (  8            partition p_591_201301 values less than (591,TO_DATE(2013-02-01, YYYY-MM-DD)),  9            partition p_591_201302 values less than (591,TO_DATE(2013-03-01, YYYY-MM-DD)), 10            partition p_591_201303 values less than (591,TO_DATE(2013-04-01, YYYY-MM-DD)), 11            partition p_591_201304 values less than (591,TO_DATE(2013-05-01, YYYY-MM-DD)), 12            partition p_591_201305 values less than (591,TO_DATE(2013-06-01, YYYY-MM-DD)), 13            partition p_591_201306 values less than (591,TO_DATE(2013-07-01, YYYY-MM-DD)), 14            partition p_591_201307 values less than (591,TO_DATE(2013-08-01, YYYY-MM-DD)), 15            partition p_591_201308 values less than (591,TO_DATE(2013-09-01, YYYY-MM-DD)), 16            partition p_591_201309 values less than (591,TO_DATE(2013-10-01, YYYY-MM-DD)), 17            partition p_591_201310 values less than (591,TO_DATE(2013-11-01, YYYY-MM-DD)), 18            partition p_591_201311 values less than (591,TO_DATE(2013-12-01, YYYY-MM-DD)), 19            partition p_591_201312 values less than (591,TO_DATE(2014-01-01, YYYY-MM-DD)), 20            partition p_591_201401 values less than (591,TO_DATE(2014-02-01, YYYY-MM-DD)), 21            partition p_591_201402 values less than (591,TO_DATE(2014-03-01, YYYY-MM-DD)), 22            partition p_591_max values less than (591,maxvalue), 23            partition p_592_201301 values less than (592,TO_DATE(2013-02-01, YYYY-MM-DD)), 24            partition p_592_201302 values less than (592,TO_DATE(2013-03-01, YYYY-MM-DD)), 25            partition p_592_201303 values less than (592,TO_DATE(2013-04-01, YYYY-MM-DD)), 26            partition p_592_201304 values less than (592,TO_DATE(2013-05-01, YYYY-MM-DD)), 27            partition p_592_201305 values less than (592,TO_DATE(2013-06-01, YYYY-MM-DD)), 28            partition p_592_201306 values less than (592,TO_DATE(2013-07-01, YYYY-MM-DD)), 29            partition p_592_201307 values less than (592,TO_DATE(2013-08-01, YYYY-MM-DD)), 30            partition p_592_201308 values less than (592,TO_DATE(2013-09-01, YYYY-MM-DD)), 31            partition p_592_201309 values less than (592,TO_DATE(2013-10-01, YYYY-MM-DD)), 32            partition p_592_201310 values less than (592,TO_DATE(2013-11-01, YYYY-MM-DD)), 33            partition p_592_201311 values less than (592,TO_DATE(2013-12-01, YYYY-MM-DD)), 34            partition p_592_201312 values less than (592,TO_DATE(2014-01-01, YYYY-MM-DD)), 35            partition p_592_201401 values less than (592,TO_DATE(2014-02-01, YYYY-MM-DD)), 36            partition p_592_201402 values less than (592,TO_DATE(2014-03-01, YYYY-MM-DD)), 37            partition p_592_max values less than (592,maxvalue), 38            partition p_593_201301 values less than (593,TO_DATE(2013-02-01, YYYY-MM-DD)), 39            partition p_593_201302 values less than (593,TO_DATE(2013-03-01, YYYY-MM-DD)), 40            partition p_593_201303 values less than (593,TO_DATE(2013-04-01, YYYY-MM-DD)), 41            partition p_593_201304 values less than (593,TO_DATE(2013-05-01, YYYY-MM-DD)), 42            partition p_593_201305 values less than (593,TO_DATE(2013-06-01, YYYY-MM-DD)), 43            partition p_593_201306 values less than (593,TO_DATE(2013-07-01, YYYY-MM-DD)), 44            partition p_593_201307 values less than (593,TO_DATE(2013-08-01, YYYY-MM-DD)), 45            partition p_593_201308 values less than (593,TO_DATE(2013-09-01, YYYY-MM-DD)), 46            partition p_593_201309 values less than (593,TO_DATE(2013-10-01, YYYY-MM-DD)), 47            partition p_593_201310 values less than (593,TO_DATE(2013-11-01, YYYY-MM-DD)), 48            partition p_593_201311 values less than (593,TO_DATE(2013-12-01, YYYY-MM-DD)), 49            partition p_593_201312 values less than (593,TO_DATE(2014-01-01, YYYY-MM-DD)), 50            partition p_593_201401 values less than (593,TO_DATE(2014-02-01, YYYY-MM-DD)), 51            partition p_593_201402 values less than (593,TO_DATE(2014-03-01, YYYY-MM-DD)), 52            partition p_593_max values less than (593,maxvalue) 53            ) 54            ; 55  56  57  58  59 --以下是插入2013年一整年日期随机数和表示福州,厦门,宁德三地的地区号含义(591到593)的随机数记录,共有10万条,如下: 60 insert into range_part_mult_col_tab (id,deal_date,area_code,nbr,contents) 61       select rownum, 62              to_date( to_char(sysdate-365,J)+TRUNC(DBMS_RANDOM.VALUE(0,365)),J), 63              ceil(dbms_random.value(591,593)), 64              ceil(dbms_random.value(18900000001,18999999999)), 65              rpad(*,400,*) 66         from dual 67       connect by rownum <= 100000; 68 commit; 69  70  71  72 select count(*) from range_part_mult_col_tab partition (p_591_201301); 73 select count(*) from range_part_mult_col_tab partition (p_591_201302); 74 select count(*) from range_part_mult_col_tab partition (p_591_201303); 75 select count(*) from range_part_mult_col_tab partition (p_591_201304); 76 select count(*) from range_part_mult_col_tab partition (p_591_201305); 77 select count(*) from range_part_mult_col_tab partition (p_591_201306); 78 select count(*) from range_part_mult_col_tab partition (p_591_201307); 79 select count(*) from range_part_mult_col_tab partition (p_591_201308); 80 select count(*) from range_part_mult_col_tab partition (p_591_201309); 81 select count(*) from range_part_mult_col_tab partition (p_591_201310); 82 select count(*) from range_part_mult_col_tab partition (p_591_201311); 83 select count(*) from range_part_mult_col_tab partition (p_591_201312); 84 select count(*) from range_part_mult_col_tab partition (p_591_max); 85 select count(*) from range_part_mult_col_tab partition (p_592_201301); 86 select count(*) from range_part_mult_col_tab partition (p_592_201302); 87 select count(*) from range_part_mult_col_tab partition (p_592_201303); 88 select count(*) from range_part_mult_col_tab partition (p_592_201304); 89 select count(*) from range_part_mult_col_tab partition (p_592_201305); 90 select count(*) from range_part_mult_col_tab partition (p_592_201306); 91 select count(*) from range_part_mult_col_tab partition (p_592_201307); 92 select count(*) from range_part_mult_col_tab partition (p_592_201308); 93 select count(*) from range_part_mult_col_tab partition (p_592_201309); 94 select count(*) from range_part_mult_col_tab partition (p_592_201310); 95 select count(*) from range_part_mult_col_tab partition (p_592_201311); 96 select count(*) from range_part_mult_col_tab partition (p_592_201312); 97 select count(*) from range_part_mult_col_tab partition (p_592_max); 98 select count(*) from range_part_mult_col_tab partition (p_593_201301); 99 select count(*) from range_part_mult_col_tab partition (p_593_201302);100 select count(*) from range_part_mult_col_tab partition (p_593_201303);101 select count(*) from range_part_mult_col_tab partition (p_593_201304);102 select count(*) from range_part_mult_col_tab partition (p_593_201305);103 select count(*) from range_part_mult_col_tab partition (p_593_201306);104 select count(*) from range_part_mult_col_tab partition (p_593_201307);105 select count(*) from range_part_mult_col_tab partition (p_593_201308);106 select count(*) from range_part_mult_col_tab partition (p_593_201309);107 select count(*) from range_part_mult_col_tab partition (p_593_201310);108 select count(*) from range_part_mult_col_tab partition (p_593_201311);109 select count(*) from range_part_mult_col_tab partition (p_593_201312);110 select count(*) from range_part_mult_col_tab partition (p_593_max);111 112 113 114 115 create index idx_part_mul_id  on range_part_mult_col_tab (id) ;116 create index idx_part_mul_nbr on range_part_mult_col_tab (nbr) local;
联合分区脚本代码

  组合分区

 1 --组合分区示例 2 drop table range_list_part_tab purge; 3 --注意,此分区为范围分区 4 create table range_list_part_tab (id number,deal_date date,area_code number,nbr number,contents varchar2(4000)) 5            partition by range (deal_date) 6              subpartition by list (area_code) 7              subpartition TEMPLATE 8              (subpartition p_591 values  (591), 9               subpartition p_592 values  (592),10               subpartition p_593 values  (593),11               subpartition p_594 values  (594),12               subpartition p_595 values  (595),13               subpartition p_596 values  (596),14               subpartition p_597 values  (597),15               subpartition p_598 values  (598),16               subpartition p_599 values  (599),17               subpartition p_other values (DEFAULT))18            ( partition p_201301 values less than (TO_DATE(2013-02-01, YYYY-MM-DD)),19              partition p_201302 values less than (TO_DATE(2013-03-01, YYYY-MM-DD)),20              partition p_201303 values less than (TO_DATE(2013-04-01, YYYY-MM-DD)),21              partition p_201304 values less than (TO_DATE(2013-05-01, YYYY-MM-DD)),22              partition p_201305 values less than (TO_DATE(2013-06-01, YYYY-MM-DD)),23              partition p_201306 values less than (TO_DATE(2013-07-01, YYYY-MM-DD)),24              partition p_201307 values less than (TO_DATE(2013-08-01, YYYY-MM-DD)),25              partition p_201308 values less than (TO_DATE(2013-09-01, YYYY-MM-DD)),26              partition p_201309 values less than (TO_DATE(2013-10-01, YYYY-MM-DD)),27              partition p_201310 values less than (TO_DATE(2013-11-01, YYYY-MM-DD)),28              partition p_201311 values less than (TO_DATE(2013-12-01, YYYY-MM-DD)),29              partition p_201312 values less than (TO_DATE(2014-01-01, YYYY-MM-DD)),30              partition p_201401 values less than (TO_DATE(2014-02-01, YYYY-MM-DD)),31              partition p_201402 values less than (TO_DATE(2014-03-01, YYYY-MM-DD)),32              partition p_max values less than (maxvalue))33            ;34 35 36 37 --以下是插入2013年一整年日期随机数和表示福建地区号含义(591到599)的随机数记录,共有10万条,如下:38 insert into range_list_part_tab(id,deal_date,area_code,nbr,contents)39       select rownum,40              to_date( to_char(sysdate-365,J)+TRUNC(DBMS_RANDOM.VALUE(0,365)),J),41              ceil(dbms_random.value(590,599)),42              ceil(dbms_random.value(18900000001,18999999999)),43              rpad(*,400,*)44         from dual45       connect by rownum <= 100000;46 commit;47 48 49 select count(*) from range_list_part_tab partition (p_591);50 select count(*) from range_list_part_tab partition (p_201302);51 select count(*) from range_list_part_tab partition (p_201303);52 select count(*) from range_list_part_tab partition (p_201304);53 select count(*) from range_list_part_tab partition (p_201305);54 select count(*) from range_list_part_tab partition (p_201306);55 select count(*) from range_list_part_tab partition (p_201307);56 select count(*) from range_list_part_tab partition (p_201308);57 select count(*) from range_list_part_tab partition (p_201309);58 select count(*) from range_list_part_tab partition (p_201310);59 select count(*) from range_list_part_tab partition (p_201311);60 select count(*) from range_list_part_tab partition (p_201312);61 select count(*) from range_list_part_tab partition (p_max);62 63 --注意,模板的形式,子分区名是被自动命名了,系统自动组合在一起,如P_201301_P_59164 select count(*) from range_list_part_tab  subpartition(P_201301_P_591);65 66 create index idx_ran_list_part_id  on range_list_part_tab (id) ;67 create index idx_ran_list_part_nbr on range_list_part_tab (nbr) local;
组合分区sql脚本示例

 

三、分区表相关信息的查询脚本

  

  该表是否是分区表,分区表的分区类型是什么,是否有子分区,分区总数有多少

SELECT partitioning_type,  subpartitioning_type,  partition_count  FROM  user_part_tables WHERE  table_name = ‘TABLE‘;

  该分区表在哪一列上建分区,有无多列联合建分区

SELECT    column_name,    object_type,    column_positionFROM    user_part_key_columnsWHERE    NAME = TABLE;

  该分区表有多大

select sum(bytes) / 1024 / 1024from user_segmentswhere segment_name =TABLE;

   该分区表各分区分别有多大,各个分区名是什么

select partition_name,        segment_type,        bytes  from user_segments where segment_name =TABLE;

  该分区表的统计信息收集情况

select table_name,       partition_name,       last_analyzed,       partition_position,             num_rows  from user_tab_statistics t where table_name =‘TABLE;

  分区表索引相关查该分区表有无索引,分别什么类型,全局索引是否失效,此外还可看统计信息收集情况。--(其中status值为N/A 表示分区索引,分区索引是否失效是在user_ind_partitions中查看)

---RANGE_PART_TAB  第一个范围分区测试脚本中直接执行下面脚本即可有相同的结果select table_name,        index_name,        last_analyzed,       blevel,       num_rows,       leaf_blocks,       distinct_keys,       status  from user_indexes where table_name =RANGE_PART_TAB;  TABLE_NAME                     INDEX_NAME                   LAST_ANALYZED  BLEVEL  NUM_ROWS LEAF_BLOCKS DISTINCT_KEYS STATUS------------------------------ ---------------------------- -------------- ------ --------- ----------- ------------- --------RANGE_PART_TAB                 IDX_PART_NBR                 01-12月-13          1    200000         536        199774 N/ARANGE_PART_TAB                 IDX_PART_ID                  01-12月-13          1    200000         555        100000 VALID--07 该分区表在哪些列上建了索引select index_name,        column_name,        column_position  from user_ind_columns where table_name = RANGE_PART_TAB; INDEX_NAME                   COLUMN_NAME          COLUMN_POSITION---------------------------- -------------------- ---------------IDX_PART_ID                  ID                                 1IDX_PART_NBR                 NBR                                1--08 该分区表上的各索引分别有多大。   select segment_name,segment_type,sum(bytes)/1024/1024  from user_segments where segment_name in       (select index_name          from user_indexes         where table_name =RANGE_PART_TAB)group by segment_name,segment_type ;  SEGMENT_NAME          SEGMENT_TYPE         SUM(BYTES)/1024/1024------------------------------------------ --------------------IDX_PART_ID           INDEX                           5IDX_PART_NBR          INDEX PARTITION            5.6875--09 该分区表的索引段的分配情况select segment_name       partition_name,        segment_type,        bytes  from user_segments where segment_name in       (select index_name          from user_indexes         where table_name =RANGE_PART_TAB);PARTITION_NAME               SEGMENT_TYPE              BYTES---------------------------- -------------------- ----------IDX_PART_ID                  INDEX                   5242880IDX_PART_NBR                 INDEX PARTITION          458752IDX_PART_NBR                 INDEX PARTITION          262144IDX_PART_NBR                 INDEX PARTITION          262144IDX_PART_NBR                 INDEX PARTITION          262144IDX_PART_NBR                 INDEX PARTITION          262144IDX_PART_NBR                 INDEX PARTITION          262144IDX_PART_NBR                 INDEX PARTITION          262144IDX_PART_NBR                 INDEX PARTITION          262144IDX_PART_NBR                 INDEX PARTITION          262144IDX_PART_NBR                 INDEX PARTITION          262144IDX_PART_NBR                 INDEX PARTITION          262144IDX_PART_NBR                 INDEX PARTITION          262144IDX_PART_NBR                 INDEX PARTITION          262144IDX_PART_NBR                 INDEX PARTITION          262144IDX_PART_NBR                 INDEX PARTITION         2097152已选择16行。                  --10 分区索引相关信息及统计信息、是否失效查看。select t2.table_name,       t1.index_name,       t1.partition_name,       t1.last_analyzed,       t1.blevel,       t1.num_rows,       t1.leaf_blocks,       t1.status         from user_ind_partitions t1, user_indexes t2where t1.index_name = t2.index_name   and t2.table_name=RANGE_PART_TAB;      TABLE_NAME        INDEX_NAME     PARTITION_NAME LAST_ANALYZED  BLEVEL  NUM_ROWS LEAF_BLOCKS STATUS-------------------------------------------------------------- ------ --------- ----------- -------RANGE_PART_TAB    IDX_PART_NBR   P_201301       01-12月-13       1        16883          45 USABLERANGE_PART_TAB    IDX_PART_NBR   P_201302       01-12月-13       1         7876          21 USABLERANGE_PART_TAB    IDX_PART_NBR   P_201303       01-12月-13       1         8448          23 USABLERANGE_PART_TAB    IDX_PART_NBR   P_201304       01-12月-13       1         8295          22 USABLERANGE_PART_TAB    IDX_PART_NBR   P_201305       01-12月-13       1         8388          23 USABLERANGE_PART_TAB    IDX_PART_NBR   P_201306       01-12月-13       1         8234          22 USABLERANGE_PART_TAB    IDX_PART_NBR   P_201307       01-12月-13       1         8540          23 USABLERANGE_PART_TAB    IDX_PART_NBR   P_201308       01-12月-13       1         8312          22 USABLERANGE_PART_TAB    IDX_PART_NBR   P_201309       01-12月-13       1         8350          23 USABLERANGE_PART_TAB    IDX_PART_NBR   P_201310       01-12月-13       1         8496          23 USABLERANGE_PART_TAB    IDX_PART_NBR   P_201311       01-12月-13       1         8178          22 USABLERANGE_PART_TAB    IDX_PART_NBR   P_201312       01-12月-13       1         8425          23 USABLERANGE_PART_TAB    IDX_PART_NBR   P_201401       01-12月-13       1         8477          23 USABLERANGE_PART_TAB    IDX_PART_NBR   P_201402       01-12月-13       1         7628          21 USABLERANGE_PART_TAB    IDX_PART_NBR   P_MAX          01-12月-13       1        75470         200 USABLE         
分区表索引相关脚本示例

 

oracle 优化 —— 分区表