Figure 9.1:
Analysis and Round-Trip
Engineering
In the current release of the HRT-UML/RCM tool-set feasibility analysis and sensitivity analysis are performed by MAST+, an enhancement of the MAST analysis utility developed at the University of Cantabria, Spain [UCA04]. In order to integrate MAST+ in the HRT-UML/RCM tool-set, we had to transform the Concurrency View to a model view which could be accepted by MAST+ (in fact by plain MAST since the MAST+ upgrade did not modify the input/output components of the original utility): what happens in practice is that a further instance of model transformation takes place from the RCM metamodel (which underpins HRT-UML/RCM) to the MAST+ metamodel and backwards.
The Analysis View is the place where the system model is described in terms of the MAST+ formalism. Although the model represented in the Analysis View need not be transformed into source code targeted for any particular platform, the information used to build it is platform-specific, and thus the Analysis View is part of PSM. The Analysis view abstracts away the details which are not relevant to its intent and purpose. For example, the system model is reduced to the instance level. Accordingly, the results of the analysis are reported back to the individual instances of the system, in every applicable view.
Once the relevant information is imported from the Concurrency View and the MAST+ model is constructed, the analysis is performed, and the Analysis View is decorated with its results and the new information automatically propagates throughout the entire model, thereby decorating the Concurrency, Interface and Deployment Views. This back propagation is part of the essential mechanisms of round-trip engineering. Some results of the analysis are placed back on individual ports and on the port clusters which are attached to VMLC instances (and then to the originating APLC); others are attached to partitions and logical connections. In the following we enumerate a few examples of the information reported back from the Analysis View:
In its current release MAST+ does not fully support
distributed
analysis as yet (but work is in progress to incorporate this feature).
Accordingly this Tutorial
cannot show and discuss the proceedings of the analysis of the
Distributed Toy Example. We
shall therefore limit ourselves to presenting the analysis of the
Partitioned Toy Example. The Tutorial will be completed with the
missing information before the end of the project.
Figure 9.2: Sample of results from the feasibility analysis
Figure 9.2 shows the results of the analysis performed on the TMTC APLC of the Partitioned Toy Example. The top of the figure shows an instance of TMTC APLC, whereas the bottom part of the figure reports the results of the timing analysis performed on port Send (which is typed Sporadic). The designer has set those values pointed to by the side arrows: the criticality level to 1; the deadline to 2,000; and the minimum inter-arrival time between any two subsequent activations of the deferred operation Send, to 10,000 units of time. The other values represent the results of the analysis. A simplified description of the measured attributes follows:
At the time this report is being written some planned features of the MAST+ analysis tool are not completed as yet. In particular: