XP 2018

TDD can give us a lot of information about functional correctness of a software system, but the way it is generally used currently cannot give much information on the performance characteristics of the implemented system. We typically do not find out about performance problems until the whole system is tested together, or worse, when it fails in production.

In this session we introduce new techniques for constructing unit tests that allow us to explore performance characteristics and detect problems before deploying the software. We can use virtual time to run performance experiments without waiting for real time to elapse, so we can get the fast feedback we are used to from the TDD cycle.

The tools and techniques present are based on adapting academic performance modelling and measurement work to a more practical context, allowing developers to harness it in their day to day development process.

We have developed a novel extension to a popular Java mock-object library that allows the specification of performance characteristics for mocked components. We can then use these to test performance characteristics without needing to assemble to whole system, or to wait for real time to elapse while tests run. This allows us to inspect performance characteristics early when we make changes to software, with quick feedback, and adapt the software appropriately, before suffering negative effects in production.