Given the complexity of the NFS environment, it is important to establish a methodology for tuning performance. The following steps provide a guideline that highlights particular areas relevant to the NFS environment. Implementing the guideline may involve more than one person or support organization.

Identify Performance Requirements: Before you begin performance tuning in general, and particularly z/OS NFS, determine those areas where performance is unsatisfactory. This is a good time to establish more precise performance requirements. As users and application requirements are identified, it can be advantageous to rank or group them according to their requirements.

Know the NFS Environment: In the previous section, the NFS environment was discussed. It is very important to fully understand the performance of the existing NFS environment, particularly that of the network. Such analysis can eliminate unnecessary tuning of the NFS client and server systems.

Establish Performance Objectives: Once the performance requirements and the NFS environment are known, you are in a position to define and prioritize your performance objectives. The performance objectives should be specified in a manner that is quantifiable and measurable. Keep in mind that you will need an executable workload or test scenario to evaluate the effectiveness of your performance tuning.

Define Workloads and Test Scenarios: You may already have workloads or test scenarios depending on how requirements and/or objectives were defined. However, these may be unwieldy or impractical to use for your performance tuning purposes. Therefore, it is advantageous to spend some time initially to define some simplified test cases that can be executed in a repeatable fashion and with as much control as is feasible. Simple test cases are also useful in diagnosing performance problems, particularly to locate an area within the NFS environment that may be impacting performance.

Select Monitoring Tools: While you may only be interested in the performance of z/OS NFS, you will find it useful to have access and familiarity with a variety of performance hardware and software monitoring tools. Not only will such knowledge assist you in collecting data to evaluate performance, but it will also help you to identify areas that may be impacting the performance of z/OS NFS. Minimally, a set of monitoring tools must be identified to collect the data upon which performance tuning decisions will be made and to determine the effectiveness of tuning.

Collect Performance Data: At this point you should be ready to begin collecting performance data. Initial measurements will be the starting point, or baseline, that will be used to evaluate the effect of your tuning. Since there can be a significant variation in network and system performance, it is prudent to repeat a performance measurement to establish the degree of variation inherent in the measurement. Doing this will provide a sense of whether or not future tuning is really affecting performance or simply normal variation.

While it may be convenient to collect performance data when systems and networks are idle, this is probably not practical. However, it is useful to collect data during peak and low activity periods. If user or application requirements are related to a specific time period, or sensitive to other system and network activity, data should be collected for these periods of time, as well. Remember that redistributing the workload may be the most cost effective approach to performance tuning.

Evaluate Performance Data: This may seem like an obvious step. However, keep in mind that the NFS environment may be quite complex. The more complex the NFS environment and your test cases are, the more data there will be to evaluate. You may also have both client and server data to evaluate as well as network data.

As you begin evaluating performance data you have collected, look for areas or opportunities where performance could be improved. Before attempting to tune z/OS NFS, you should determine if performance is primarily impacted by the NFS server system, the NFS client system, or the network itself. In fact, you may determine that additional data must be collected prior to any performance tuning. You may also discover evidence that configurations and parameter settings within the evaluation environment are not optimal.

The output of this step is a list of changes that you believe will positively affect performance. As you make this list, identify the impact or cost associated with a change. It is also useful to identify any resources that are heavily utilized or that have contending requirements. This will help you to prioritize and ultimately select the changes you will make or recommend.

Tune Your NFS Environment: Make one change from your list of possible tuning changes at a time. Measure and evaluate the effect of that change before making any other changes. Pay particular attention to any impact on heavily utilized resources that may have been previously identified in addition to newly exposed resources that now appear to be impacted. Since performance tuning typically involves a trade off in resource utilization, make sure you have not inadvertently caused a performance problem elsewhere. Also, before deciding to implement a change, consider whether any observed changes are due to the tuning change you've made or simply a result of normal measurement variation. Repeat this step as necessary.