Performance Testing and Performance Modeling

Many organizations are enforcing good performance testing activities. These include defining the goals of the test, characterizing the behavior of virtual users, and thoughtful establishment of a performance test environment and data. For complex environments there are typically two challenges – the cost of sizing an acceptable environment and test scheduling conflicts with end-to-end components such as a mainframe test environment.Performance testing delivers some key information:

Performance testing can help with identification of bottlenecks of the application in the test environment

• It can help identify memory leaks during endurance tests
• It can help identify tuning attributes of the test environment that may be applicable to the production environment
• It can be very valuable to help identify the relationships between transactions and resource consumption to build a performance model

Performance modeling can extend the value of performance testing:

It can translate information found in the test environment to results that will occur in the production environment. Because modeling simulations can be done quickly to test a variety of questions, it can offer a broader spectrum of performance and cost trade-off answers in a shorter timeframe and at a lower cost than performance testing by itself. Changes to a model can be done incrementally. This supports a variety of performance question strategies from a top-down approach to a system by system approach.

Simulation modeling can extend performance testing and provide an accurate answer to performance and capacity requirements in production - avoiding over capitalizing the production environment or the test environment. The ROI is real and easy to measure.

Standard Performance Modeling Approach

The standard approach to gathering performance measures to build a performance simulation model is to leverage system resource measures that are native and correlate them with the workload of transactions of a performance test. In many cases, Genilogix leverages the integrated monitors of Mercury LoadRunner to gather this data.

Questions that can be answered with modeling

1. What will the response time and resource consumption for a given number of business functions in the production environment after transport or go-live?
2. What improvement would be seen in production by varying the type of hardware, the number of servers or the network topology?
3. Would consolidation of sub-systems onto a centralized instance impact performance?
4. What would be the expected impact in production if the number of users or the ratio of transactions changed?

Performance Modeling Methodology

At Genilogix, we use the HyPerformix IPS Performance Optimizer (Mercury Capacity Planning) and Mercury LoadRunner to enable performance modeling. A standard detailed methodology is followed to deliver rapid results form performance modeling engagements. This methodology includes a five step approach:

1. Define goals, objectives, and data requirements
2. Gather data
3. Build, verify and validate the model
4. Predict the performance and capacity utilization in production
5. Evaluate alternative scenarios