This process involves the following key aspects:
Step 1:
Identify and prioritize business use scenarios and applicable performance scenarios, which should be mapped to each of the business-critical processes. The scenarios should be compiled based on interviews with the business and by analyzing historical data. Next, the scenarios should be prioritized based upon their importance and occurrence probabilities. Each of these scenarios should be mapped to a design principle that should be implemented.
For example, in an e-commerce application, the shopping cart steps and checkout process are the most critical scenarios, and they directly contribute to revenues. Product recommendations, search experience, and personalization scenarios provide an indirect contribution to revenue.
Step 2:
Model the load for each of the business scenarios identified in the previous step, and identify the potential workload. The workload could be one of the following:
- Total number of users (anonymous and registered) and the average think time
- Number of concurrent users
- Number of page visits
- Number of transactions per hour
- The volume of input and output data
Response time objectives are substantial in the context of a defined organizational workload. It may be easier for an application to process a transaction in six seconds, but difficult to realize this goal when receiving 1,000 transactions per second. This means that the requirement needs to specify the context of a specified goal, consisting of when a transaction begins and when it finishes. In many applications, response time under a fixed load will vary according to the distribution curve. Most transactions will complete close to the average response time mark, but some will take longer, and a few will take less time. Usually, it's unreasonable to expect all transactions to complete within a target response time. It is more realistic to assume a proportion, such as 80- 90%, to meet these objectives.
Throughput is defined in terms of transactions per second, or minute, or hour, where a transaction is a measurable unit of work. The transactions leveraged for throughput planning should typically be established from the application's primary business scenarios. Workload forms a critical input in designing the enterprise application and the infrastructure. Use surveys, historical analysis of the existing systems, and a business roadmap all provide input to the workflow modeling process.
Step 3:
Identify performance-related patterns after determining the key scenarios, and model the workload for each of them. The following are a few instances of patterns, similar trends related to locations, access devices, and demographics should be analyzed. This helps in accurate performance modeling and testing. This modeling contributes to quantify the performance expectations and helps in accurately design.