The advantages of parallel and distributed software systems in terms of additional reliability, redundancy, work load balancing etc. are easily outweighed by the additional complexity parallelism and distribution introduce into a software architecture. In this paper we consider an approach to describe the architecture of parallel and distributed software systems. This approach is based on a component model of software which contains special constructs for concurrency control and additional information about distribution. Rather than describing the distribution properties within a component most of these properties are stated with the use relation between components which may be local or remote. We describe how this design approach can be implemented on top of CORBA and how performance-related properties of remote use relations are used to quantitatively assess the software architecture. Thus the design of complex, hierarchically structured distributed software systems making full use of parallelism can be assessed wrt. response time of remote operation invocations, for example.