Early Design Phase Cross-Platform Throughput Prediction for Industrial Stream-Processing Applications

To architect and design a system, the stakeholder needs have to be satisfied by technical solutions, for which decisions on trade-offs have to be made. A trend is that the number of functions, components, and interfaces in systems increase, often by an order of magnitude or more, such that reasoning about the impact of a decision becomes increasingly hard and tracing its impact throughout the system is crucial. Therefore, we decompose a system in areas of knowledge and information, which we call knowledge domains. Architecting means taking decisions, for which the impact on knowledge domains and their explicit relations are required. Existing approaches that reason across systems either do not make explicit relations between knowledge domains, or perform a quantitative computation instead of reasoning, where for both it is difficult to trace the impact of decisions. In this paper, we present an architecture reasoning structure with which knowledge domains from different disciplines can be explicitly related, which enables system wide reasoning and decision making. With just eight language elements, a system can be described in an information structure. By applying a knowledge domain pattern, the essential information of knowledge domains is captured. Via relations, both qualitative and quantitative reasoning can be performed to trace the impact of decisions. An example is used to illustrate the approach, for which the tension for a decision is shown by tracing its impact via quantitative and qualitative relations. The approach was investigated and validated in the industrial context of Oce professional printing systems.