一个分区表上的GLOBAL索引和普通索引相比有什么优势？

Yong Huang

原帖由 translate 于 2010-1-25 00:17 发表
create index t_idx
on t(owner,object_type,object_name)
global
partition by hash(owner)
partitions 16

If you hash-partition the index on owner, you'll have much less contention on index root block for high concurrent queries like "... where owner = 'ABC'". A non-partitioned index can't sustain high concurrency due to root block contention.

Yong Huang

translate

原帖由 Yong Huang 于 2010-1-26 03:55 发表


If you hash-partition the index on owner, you'll have much less contention on index root block for high concurrent queries like "... where owner = 'ABC'". A non-partitioned index can't sustain high concurrency due to root block contention.

Yong Huang

    谢谢

      这个只是一个例子，在TOM这本书里面讲了大量的LOCAL PARTITION索引和GLOBAL PARTITION索引的区别，但是我想知道什么时候在一个PARTITION表上如果建立一个普通索引和建立GLOBAL PARTITION索引有什么区别

Yong Huang

原帖由 translate 于 2010-1-27 03:18 发表
      这个只是一个例子，在TOM这本书里面讲了大量的LOCAL PARTITION索引和GLOBAL PARTITION索引的区别，但是我想知道什么时候在一个PARTITION表上如果建立一个普通索引和建立GLOBAL PARTITION索引有什么区别

"区别" in what sense? I already said there's difference in reducing contention. Do you want to know the difference in administration? They both become unusable after either the whole table or one of the partitions goes through certain DDL. Difference in syntax to create them? You already know it.

Yong Huang