High-capability software services, like transaction processing systems, need to satisfy a range of non-functional characteristics such as performance, availability, and scalability. To fulfill these needs, the core business logic is usually extended with a large amount of non-domain logic in the form of frameworks, libraries, and custom code, which sometimes cannot be cleanly separated from the domain logic. So, it is nearly impossible to generate test cases for the whole system systematically guided by structural metrics on the source code. In this paper, we propose a specification-based approach to generate test cases. In this approach, the domain logic is specified in a functional notation (based on Gallina). Test cases are generated using a search-based approach where the fitness function is defined in terms of the structural coverage of the specification (measured over an equivalent Haskell implementation). An experiment on an industrial stock exchange trading engine indicates promising results in the effectiveness of our proposed approach.