You allow WLM to manage the number of servers and server address
spaces for an application environment if you choose the option “Managed
by WLM” for starting servers of a subsystem instance. In this case,
WLM has the ability to consider the delays of work requests sent to
the application environments in its algorithms to supervise the goal
achievement of your service classes and to adjust the resources needed
by them. Servers and server address spaces are considered resources
which can be made available to the work using an application environment.
This is illustrated in the following example:
- If a DB2® stored procedure
request comes into the system, DB2 classifies
the work request to WLM and WLM assigns a service class to it. DB2 then queues the work request
to the application environment under which the stored procedure should
be executed.
- WLM queues the work requests for each application environment
by service class. This allows WLM to understand how the queuing affects
the goal achievement of the service class.
- When the first request is queued to an application environment,
workload manager detects that there are no active servers for the
request, and automatically starts one.
- From then on, WLM collects statistics about the queue delays for
each application environment and each service class used for the work
requests. These queue delays then become part of the WLM algorithms
which assess the goal achievement of the service classes and adjust
the resources as needed.
- If, for example, the service class for the DB2 stored procedure requests do not meet their
goals, WLM determines which resources are needed to help the work
in the service class.
- If queue delay shows up as the dominating factor, WLM assesses
how many additional servers are needed to help the work to meet their
goals or at least to see a significant improvement.
- Once the number of servers has been assessed, WLM makes sure that
the system resources are available to start the necessary server address
spaces for the required number of servers. During this step, WLM makes
sure that more important work is not affected and that the system
does not run into a shortage situation because of the new server address
spaces. Among the resources considered are:
- Processor
- Real storage
- Auxiliary storage constraints
- Common storage (SQA) constraints
- When all tests have completed successfully, WLM starts the required
number of server address spaces.
In cases where the system has low utilization, WLM is also able
to start one additional server address space for the application environment
if this may immediately help the work and it is granted that sufficient
resources remain available for other work in the system.
Server address spaces are stopped when the utilization of the servers
drops and many servers become idle. WLM then returns the resources
used by the server address spaces and thus allows other work to utilize
the system.
Using “Managed by WLM” is the optimal way to manage the number
of server address spaces. It provides the best performance for the
work executed under the application environment and only uses the
resources which are optimal based on the goal for the work and the
overall utilization of the system.