Improving Accuracy and Confidence in Workload Models
The most critical component in capacity planning and performance engineering is the Workload Model, which defines the workflows, throughputs, and target performance your system must support at peak loads. As critical as it is, it can be difficult and particularly challenging to predict loads for new applications, features, or events. A typical approach starts with a wild-guess worst-case scenario—but overestimates waste time and money, and force you to engineer applications and infrastructure to support unrealistic loads. Low estimates can result in terrible customer experiences, lost revenue, and costly remediation. So you need realistic numbers about which you are confident. Gopal Brugalette and Safi Mohamed share a set of techniques to develop more accurate models. They discuss the potential gaps in a model arising from differences between customer behavior and system design. Gopal and Safi also discuss advanced approaches such as Fermi estimates, statistical forecasting, and ways to validate assumptions and predictions. Learn these techniques to improve your workload models and increase your confidence.