%%[[Load Testing]] [[Performance Testing]] [[Statistics]] [[Data science]] [[Load testing is a data science]] [[Analyzing load testing results]]%% %% date:: [[2023-03-13]] %% # [[Sample sizes for load testing]] In load testing, sample sizes are important on two levels: - The sample size of the production environment's historical metrics (response time, transaction distribution, user concurrency) must be large enough to describe the production population - The sample size of a single test must be large enough to be able to describe the characteristics of the test environment population Only if these two conditions are met can the results of a test accurately represent what will happen in production. In the Planning phase, a large enough sample must be taken from production monitoring services to inform the [[Workload Modeling]]. [[How to build a workload model]] In the Results Analysis phase, a large enough sample size should be taken (ie., enough tests should have been executed) so as to make conclusions about the test environment. Then, metrics like response time, user concurrency, throughput, transaction distribution, etc., from the test environment should be compared to those of the production environment. For the test to be a valid simulation of production, they should be sufficiently similar. "Calculadora del Tamaño Muestral (https://raosoft.com/samplesize.html) [[Calculating sample size required for specified confidence interval]]" - "Confidence level" - "dice cuántos usuarios o muestras necesitamos con un intervalo de confianza de 90%, 95%, 99%, etc." - "Tolerancia (margen de error)" - "intervalo de confianza - nivel de incertidumbre" - "tamaño de la población" - "distribución de la respuesta"