Most event-driven systems allow a huge number of possible event sequences, which makes exhaustive testing infeasible. As a possible solution, usage-based testing has been proposed for several types of event-driven software.

I. INTRODUCTION

• An example of event-driven software is the Service Oriented Architecture (SOA), where service providers and service consumers communicate with each other using events
• One approach to resolve this scalability problem is usage-based testing, a goal oriented testing methodology that prioritizes testing effort based on how often which functionality of the system is used
• The system’s usage is described by probabilistic usage profiles, e.g., Markov chains that describe the probability of the next event.
• To obtain these profiles, usage mining is used, i.e., analyzing logged executions of the system’s usage
• The usage-based testing model model is divided into three layers:

• the platform layer - that depend on the platform on which the System Under Test (SUT) is implemented,
• the translation layer - the translation layer translates the platform specific events into abstract events that only consist of string representation of both the event type and event target.
• the event layer - r is platform independent. It works on abstract events and provides functionality for the training of probabilistic usage profiles, analysis of these profiles, and testing task, like test case generation

II. MODEL OUTLINE

Platform layer

It contains the platform specific parts of the model, i.e., a recorder to monitor the SUT and log its execution and a player that can send stimuli to the SUT to replay event patterns.