There are three rules to performance optimizations: Don't optimize yet!, Don't optimize yet!, Don't optimize yet! Traditionally, the popular notion towards performance has been to ignore it until it could no longer be ignored. However, many companies have faced the consequence of this approach, finding that some performance problems are impossible to solve through code optimization alone. A discipline called Software Performance Engineering(SPE) tries to alleviate performance problems before they arise by modelling performance already at the design phase. However, SPE has never caught on in the mainstream Software Engineering practices - mainly because it relies on specialized performance staff. In this thesis we propose an approach making early performance considerations more amenable to Software Engineering. We have created a J2ME test application that derives performance characteristics of the target environment. The test application can be used to supply input estimates that would otherwise need to be supplied by a SPE specialist. The input estimates from the test application are used to annotate activity diagrams according to the UML Profile for Schedulability, Performance, and Time which results in performance predictions of the system under development. We have investigated the validity of our approach by applying it to a test case, a J2ME application.