客户有个12c的环境，正在做xtts的迁移，从10g迁移到12c。其中一项要求是12c的一些初始化参数和10g一致，如parallel_max_servers。发现设置后，重启了数据库，还是无法达到预期的修改值。

SQL> show parameter parallel_max_servers
NAME                    TYPE       VALUE
----------------------  -----     ---------
parallel_max_servers    integer    474


SQL> alter system set parallel_max_servers=585 scope=spfile;


shutdown immediate;
startup;


SQL> show parameter parallel_max_servers
NAME                    TYPE       VALUE
----------------------  -----     ---------
parallel_max_servers    integer    474

打开其alertlog中，我们可以看到类似如下的信息：

Adjusting the altered value of parameter parallel_max_servers
from 585 to 474 due to the value of parameter processes (600)

我们在Doc ID 1968840.1处看到：

From 11gR2 onwards, there is a new method to compute the default for PARALLEL_MAX_SERVERS.  

In 11gR2, the value of  PARALLEL_MAX_SERVERS is capped by PROCESSES - 15.  

In 12c, the value of PARALLEL_MAX_SERVERS is capped by PROCESSES - N, where N is the result of an internal calculation that estimates the maximum number of background processes that need to be reserved for the particular database.

parallel_max_servers的最大值，受限制与process，在11gR2中，这个限制是parallel_max_servers=processes-15（另外，在11.2.0.4中变成了30）.

而在12c中，是processes-N，这个N是一个内部算法。具体的值可以通过如下来得到：

SQL> select KSBNEEDED_VAL into v_process_reserved from sys.X$KSBDPNEEDED ;
SQL>

所以在12c中的parallel_max_servers可以设置的最大值，为process-(KSBNEEDED_VAL.X$KSBDPNEEDED)。

参考：
How PARALLEL_MAX_SERVERS and PROCESSES Parameters are Related? (Doc ID 1968840.1)
New Default Value of PARALLEL_MAX_SERVERS in Release 11.2.0.2 (Doc ID 1377411.1)
Oracle Community:The value of parallel_max_server adjusted in alertlog of 12c(12.1.0.2) database is not as per the formula mentioned in the (Doc ID 1968840.1)