Chapter 15. Performance and Capacity Planning

The decision to run a particular set of work on the mainframe entails many factors, one of which is the performance of the environment. Part of the value you can get from the Linux-on-the-mainframe environment is the efficiency of the mainframe to support the diverse workload of a server hardware consolidation environment. Since we expect the majority of the Linux-on-the-mainframe deployments to be on z/VM, this chapter focuses on performance and capacity planning in this environment.

Sometimes, performance and capacity planning on a physical server farm means watching for response time problems. When response times reach the limit of acceptability, one solution might be to split the workload and buy a new machine, complete with more CPU, memory, I/O, and cables.

Linux on the mainframe allows you to rethink your methods for performance and capacity planning. The tasks included in the performance and capacity planning discipline are listed in Figure 15-1. Tools are available on the mainframe for these tasks, both with z/VM and the Linux mainframe distributions.

Figure 15-1. What performance covers

For the simpler Linux-on-the-mainframe start-up projects such as domain name server (DNS) or static Web page serving, the focus is primarily on getting familiar with the Linux-on-the-mainframe environment. Performance issues will typically be of secondary importance in a server farm environment. When consolidating servers on the mainframe, planning for performance enables you to make efficient use of the mainframe resources. For more complex projects, for example, a business intelligence application like StoreCompany's OaK (Appendix B, "StoreCompany"), you might want to put in the necessary effort when you set up your system to ensure that you meet your performance goals.

The tools and techniques that performance experts need are readily available, and z/VM gives you great flexibility in achieving performance objectives.

In this chapter, we explore the following questions about Linux on the mainframe:

• How are performance data collected?

• How are hardware tuning and software tuning performed?

• How is capacity planning performed?

• What is unique about performance in this environment?