120 Problem Determination for WebSphere for z/OS
For a transaction to complete, it must be able to access all the resources it requires. For a
transaction to perform well, there must be enough of these resources available and they need
to be available quickly enough. How much is enough? How quick is quick? There is no firm
answer. It depends on your business requirements.
13.2 Managing performance of WebSphere transactions
WebSphere for z/OS uses the z/OS Workload Management (WLM) function to start and
manage servers in response to workload activity. It is a WebSphere for z/OS requirement that
WLM run in goal mode. If your z/OS system is running WLM in compatibility mode, follow the
instructions in z/OS Planning: Workload Management, SA22-7602, to implement the goal
mode before proceeding with WebSphere for z/OS customization.
Each J2EE application server in a WebSphere for z/OS cell uses WLM to start servants as
WLM application environments. Thus, each application server must be associated with a
WLM application environment name. The cluster transition name in the WebSphere for
z/OS configuration is used as the WLM application environment name.
WebSphere for z/OS makes use of dynamic WLM application environments when they are
available. (The WLM service that added dynamic application environments is a prerequisite
for WebSphere for z/OS.) Therefore, you do not need to use the WLM ISPF panels to
manually create application environments for WebSphere for z/OS.
The response time and throughput of WebSphere transactions are managed based on their
assigned service class, associated performance objectives, and availability of system
resources, which we explain in more detail in the following sections.
13.2.1 Managing the number of servant regions
Each WebSphere application server can have one or multiple servant regions per server
instance based upon the settings that are defined in the WLM application environment.
How many servant regions are created depends on the WLM determination of:
How the work is meeting the performance goals
The importance of the work compared to other work in the system
The availability of system resources that are needed to satisfy those objectives
Whether starting more address spaces can help achieve the objectives (see Figure 13-1
on page 121)
Chapter 13. WebSphere for z/OS performance analysis 121
Figure 13-1 WebSphere workload definition with Workload Manager
By default, the minimum number of regions for J2EE servers is one; there is no default
maximum. You can override the maximum and minimum number of servant regions that WLM
will start with three parameters in the Administrative Console:
Minimum number of instances: wlm_minimumSRCount
This parameter is used to start up a basic number of servant regions before the day's work
arrives. This can reduce the time that is spent waiting for WLM to determine that more
servant regions are needed. To keep work from coming in through the protocol handler
before servant regions are ready, use:
Maximum number of instances: wlm_maximumSRCount
This parameter is used to cap the number of address spaces started by WLM if you
determine that excessive servant regions might contribute to service degradation (for
example, if real storage is limited).
Multiple instances enabled
This parameter is used to limit application server to one servant region. Even if minimum
and maximum numbers of servant regions are defined as
> 1, the ports will not be open.
Ensure that you have selected the option to allow more servant regions.
WebSphere Server Instance
TransClassMap edgeplex.itso.ibm.com:* /webap1/myservlet WASHI
TransClassMap wtsc48oe.itso.ibm.com:* /webap2/* WASHI
TransClassMap haplex1.itso.ibm.com:7080 * WASHI
TransClassMap *:7070* /trade/* WASDF
TransClassMap * /myservlet WASLO