For any system, in particular those which are highly complex or real-time, it is essential that designers are confident that the system is viable before a prototype is built. Traditionally, this confidence has been gained by simulating the behaviour of either all or just the key parts of the system. Inviable aspects of the design can be corrected early before incurring the expense of building the system.
Whilst beneficial, such an approach is itself too late in the lifecycle. Performance should be designed into the system from the outset, not simply verified at the end. A need exists for a performance analysis capability that can be used from the earliest considerations of system architecture through to the details of system components. At each stage, performance analysis will calculate quantitative performance data used both to verify the validity and integrity of the system at the current level of design, and to provide performance design criteria for the next level of design. At each stage, if the next level of design meets its performance criteria, the system remains viable.
The performance modelling module in the Cradle environment exists to satisfy this need.
Performance modelling is an activity in which the performance of a system is characterised by a set of performance parameters whose quantitative values are used to assess the system's viability. It starts at the highest design level (often termed architecture modelling) and continues until the design components are defined, when simulation is used.
Performance modelling analyses the characteristics of the behaviour of the system and derives performance directly. Compare this approach with simulation which analyses the behaviour of the system and collects statistics from which performance is derived.
Performance modelling annotates the elements of the design model (processes, flows, equipments, buses, and links) with potentially infinite numbers of performance parameters (PPs) each defined by:
The annotation is held in the PERFormance DATA frame of each symbol in the design diagrams and creates a mathematical model to which an environmental load is applied during an analysis. Each analysis can apply to all of a subset of the model. Named subsets can be created, termed state models.
Environment loads can be applied in an analysis, defined by a three layer structure consisting of:
The results of each analysis are held separately and can be selectively reported by threads (each a named path through a state model).
Viability of the system is typically shown by the results of an analysis not violating constraints. Such results form constraints for the next level of design; if the next level performs within the constraints then the system remains viable.
Performance modelling is general purpose and can calculate costs, weights, fluid flows, or power consumptions, or whatever subject matter is represented in PP derivations.
Performance Modelling answers the 'How' Questions:
It allows the analysis of system performance very early in a project lifecycle and a clear and precise method for the definition of performance requirements which must then be met by individual sub-systems.
Requirements Management |
Systems Modelling | Performance
Modelling | Software Engineering
Document Composer | The
Cradle Core | 3SL | Home