Publishing Engine Programmer's Guide > The Arbortext Publishing Engine Request Manager > Understanding the Internal Structure of Arbortext PE Request Manager > Arbortext PE sub-process Pool Attributes > The minSubprocesses and maxSubprocesses Attributes
  
The minSubprocesses and maxSubprocesses Attributes
These attributes define the number of Arbortext PE sub-processes running in a pool. The minSubprocesses attribute sets the minimum number of Arbortext PE sub-processes that will be started and continue to run. The maxSubprocesses attribute sets the maximum number of Arbortext PE sub-processes that can run in a pool when conditions controlled by other configuration settings are met.
When an Arbortext PE sub-process pool initializes, it starts Arbortext PE sub-processes until the minimum number have been started. If a pool terminates an Arbortext PE sub-process and causes the number running to drop below the minimum number configured (refer to The maxIdleInterval Attribute and The maxLifetime Attribute’ for information on configuring terminations), the pool will immediately start a new Arbortext PE sub-process.
The maximum number of Arbortext PE sub-processes must be greater than the minimum number to start additional Arbortext PE sub-processes if needed to satisfy requests. The pool can't exceed the maximum number of Arbortext PE sub-processes, which counts both allocated and idle Arbortext PE sub-processes in the pool.
The default value for both minSubprocesses and maxSubprocesses is 1. Specifying a value less than zero or setting maxSubprocesses less than minSubprocesses is an error.
Ordinarily, set the minimum and maximum numbers of Arbortext PE sub-processes to the same value to optimize performance. An extra Arbortext PE sub-process consumes a certain amount of memory, but even if it is seldom or never asked to service a request, it's a small penalty. If you specify starting a small number of Arbortext PE sub-processes and allocate more as needed, consider the overhead associated with startup and initialization that can delay fulfilling a request.