V$PX_PROCESS_SYSSTAT contains information about the sessions running parallel execution.
Name of the statistic:
Servers In Use - Number of PX servers currently performing parallel operations
Servers Available - Number of PX servers available to perform parallel operations
Servers Started - Number of times the system has had to create a PX server process
Servers Shutdown - Number of times a PX server process has been shut down. A PX server process will be shut down if it has not been used recently.
If this value is large, then consider increasing the parameter. This will improve performance by avoiding the latency of PX server process creation.
Servers HWM - Maximum number of concurrent PX server processes
If this number is equal to the PARALLEL_MAX_SERVERS initialization parameter, then consider increasing the parameter. This could allow you to increase your throughput, especially if your system is under-utilized and the V$SYSSTAT statistic "Parallel operations downgraded to serial" is large.
Servers Cleaned Up - Number of times PMON had to clean up a PX server. This should only happen during abnormal termination of a parallel operation.
If this number is large, then you should determine the cause.
Server Sessions - Total number of sessions created by all PX servers
Memory Chunks Allocated - Number of large memory chunks allocated by PX servers
Memory Chunks Freed - Number of large memory chunks freed
Memory Chunks Current - Number of large memory chunks currently being used
Memory Chunks HWM - Maximum number of concurrently allocated chunks
Buffers allocated - Number of times a message buffer has been allocated
Buffers freed - Number of times a message buffer has been freed
Buffers Current - Number of message buffers currently being used
Buffers HWM - Maximum number of concurrently allocated message buffers
Value of the statistic
The ID of the container to which the data pertains. Possible values include:
0: This value is used for rows containing data that pertain to the entire CDB. This value is also used for rows in non-CDBs.
1: This value is used for rows containing data that pertain to only the root
n: Where n is the applicable container ID for the rows containing data