This chapter describes the optimization and tuning of the POWER7 processor-based server running WebSphere Application Server.

This chapter is intended to provide you with performance and functional considerations for running WebSphere Application Server middleware on Power Systems. It primarily describes POWER7. Even though WebSphere Application Server is designed to run on many operating systems and platforms, some specific capabilities of Power Systems are used by WebSphere Application Server as a part of platform optimization efforts.

The intent of this chapter is to explain WebSphere Application Server installation, deployment, and migration topics when WebSphere Application Server is running on Power Systems. This chapter also describes preferred practices for performance when you run enterprise Java applications on Power Systems. This chapter also highlights some of the known WebSphere Application Server topics and solutions for Power Systems.

As there are multiple versions of WebSphere Application Server, there are also multiple versions of AIX supported by POWER7. Table 9-1 shows some of the
installation considerations.

When you start the WebSphere Application Server, there is an option to bind the Java processors to specific CPU processor cores to circumvent the operating system scheduler to send the work to available processors in the pool. In certain cases, this action improves the performance. Table 9-2 lists some of the deployment considerations.

When you run WebSphere Application Server on IBM POWER7 Systems, the end-to-end performance depends on many subsystems. The subsystems include network, memory, disk, and CPU subsystems of POWER7; a crucial consideration is Java configuration and tuning. Topology also plays a major role in the performance of the enterprise application that is being deployed. The architecture of the application must be considered when you determine the best deployment topology. Table 9-3 includes links to preferred practices documents, which target each of these major areas.

Certain applications that run on WebSphere Application Server on POWER7 can experience performance degradation because of asynchronous I/O (AIO). AIO can be disabled to improve the performance of these applications. For instructions about how to accomplish this task, see Disabling AIO (Asynchronous Input/Output) native transport in WebSphere Application Server, available at:

By default, POWER7 runs in SMT4 mode. As such, there are four hardware threads (or four logical CPUs) per core that provide tremendous concurrency for applications. If the enterprise applications are migrated to POWER7 from an earlier version of POWER hardware (POWER5 or POWER6), you might experience scalability issues because the default SMT mode on POWER7 is SMT4, but on POWER5 and POWER6, the default is SMT and SMT2 mode. As some of these applications might not be designed for the massive parallelism of POWER7, performance and scalability can be improved by using smaller partitions or processor binding. Processor binding is described in “Processor affinity benefits for WebSphere applications” on page 150.

When an application is running on top of WebSphere Application Server is deployed on a large LPAR, it might not use all the cores in that LPAR, resulting in less than optimum application performance. If this situation occurs, performance improvements to these applications can be obtained by binding the application server to certain cores. This task can be accomplished by creating the resource sets and attaching them to the application server that is running excerset. For an example of using the taskset and numactl commands in a Linux environment, see “Partition sizes and affinity” on page 14.

In addition to the processor affinity described in “Processor affinity benefits for WebSphere applications”, applications can benefit from avoiding remote memory accesses by setting the environment variable MEMORY_AFFINITY to MCM. This variable allocates application private and shared memory from processor local memory.

These three tuning techniques (SMT scalability, CPU affinity, and memory affinity) can improve the performance of WebSphere Application Server on POWER7 Systems. For an example of using the taskset and numactl commands in a Linux environment, see “Partition sizes and affinity” on page 14.

More information about these topics is in Java Performance on POWER7 - Best practices, found at:

Resources for addressing issues regarding performance analysis, problem determination, and diagnostic tests are listed in Table 9-4.