Maclean Liu(刘相兵

但是我测试发现， 并行却开不起了

==>你怎么得到这个结论的？

Maclean Liu(刘相兵

PARALLEL_INSTANCE_GROUP is an Oracle RAC parameter that you can specify in parallel mode only. Used in conjunction with services or with the INSTANCE_GROUPS parameter, it lets you restrict parallel query operations to a limited number of instances. Note that the INSTANCE_GROUPS parameter has been deprecated.

This parameter identifies the parallel instance group Oracle will use for spawning parallel execution processes. If used in conjunction with services, then parallel operations will spawn parallel execution processes only on instances defined in the service. If used in conjunction with INSTANCE_GROUPS, then parallel operations will spawn parallel execution processes only on instances that specify a matching group in their INSTANCE_GROUPS parameter.

If the value assigned to PARALLEL_INSTANCE_GROUP is the name of a service or group that does not exist, then the operation runs serially. No parallelism is used.

Maclean Liu(刘相兵

并行跨实例运行
建议一般情况下并行查询限制在实例内，这可以通过INSTANCE_GROUPS、PARALLEL_INSTANCE_GROUPS参数来实现.
比如在节点1上设置：
rac1.Instance_groups=rac1,rac
Rac1.Parallel_instance_group=rac1
在节点2上设置：
Rac2.Instance_groups=rac2,rac
Rac2.Parallel_instance_group=rac2
通过这些参数的设置，在缺省环境下，并行查询只能在实例内进行。因为每个实例的缺省PARALLEL_INSTANCE_GROUP都是实例名，只有本实例的instance_groups参数中包含和本实例名相同的instance_group。而如果我们想要让某个SQL跨实例做并行查询的时候，可以通过下面的方法进行：
Alter session set parallel_instance_group='rac';
Select ....
由于两个实例都属于rac instance_group，所以只要parallel_instance_group设置为‘rac’就可以了。