Performance engineering has become a central plank in the design of complex, time-critical systems. It is supported by stochastic modelling, a brief history of which is given, going back to Erlang as long ago as 1909. This in turn developed according to successive new generations of communication and computer architectures and other operational systems. Its evolution through queues and networks is reviewed, culminating in the unification of many specification and solution techniques in a common formalism, stochastic process algebra. Recent results are given on the automatic computation of separable solutions for the equilibrium state probabilities in systems specified in such a formalism. A performance engineering support environment is proposed to integrate these methods with others such as response time analysis and fluid models, which are better suited to large scale aggregation of similar components in a continuous space.