A Rationale-based Model for Architecture Design Reasoning
A Rationale-based Model for Architecture Design Reasoning
A Rationale-based Model for Architecture Design Reasoning
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
10.3. <strong>Reasoning</strong> about change impact with AREL<br />
is most influential because if we change the fundamental design of asynchronous processing,<br />
say to synchronous, then the assumption of how to detect loss of acknowledgement<br />
messages would change and the subsequent design would also be different. This is reflected<br />
in the CPT of AR14 where the probability of C4 2 4 being volatile whilst the other requirements<br />
being stable is 0.3. C4 2 4 also affects decisions AR13 and AR16, their CPTs<br />
are shown in Figure 10.5(b).<br />
Figure 10.5(b) shows the prior probabilities and the CPTs of the security design <strong>for</strong><br />
payment message processing. There are a number of key decisions, AR13 is a decision on<br />
the processing steps of MCP. It is influenced by C4 2 17 because it must be decrypted<br />
be<strong>for</strong>e the message can be processed. It is influenced by C4 2 5 because the message<br />
must then be validated. If validation fails, the message is sent back to the originator with<br />
a negative acknowledgement immediately. The decision is influenced by C4 2 4 because<br />
this decision is <strong>based</strong> on the assumption that the treatment of acknowledgement is asynchronous,<br />
there<strong>for</strong>e MCP must be stateful, i.e. MCP has to remember the status of a<br />
payment message so that when the acknowledgement is returned at some point in the<br />
future, it could be matched to the original message.<br />
AR16 is a decision on processing acknowledgement. It depends on a number of factors<br />
including C4 2 4, C4 2 5 and C4 2 17. The decision relies heavily on C4 2 4 in that if<br />
the acknowledgement processing model changes, it is highly likely that the decision will<br />
become invalid.<br />
10.3.2 Original beliefs modelled by AREL<br />
The AREL model in Figure 10.6 is a representation combining the illustrations in Figure<br />
10.5(a) and 10.5(b) without the details of the prior probabilities and CPTs. It shows the<br />
marginal probability distribution when no evidence (i.e. change) has been added. The<br />
specific visualisation <strong>for</strong>mat is provided by Netica [107] BBN software. The visualsation<br />
includes the name of each node across the top, the state names are on the left hand side<br />
of the nodes (valid/invalid <strong>for</strong> AR nodes and stable/volatile <strong>for</strong> AE nodes). Marginal<br />
probabilities are shown as percentages (e.g. 85.6% and 14.4% <strong>for</strong> AR10 ) by the two<br />
horizontal bars. The marginal probabilities, <strong>for</strong> instance, of a leaf-node such as C4 2 9,<br />
are calculated <strong>based</strong> on its own CPT and the CPTs and prior probabilities distributions<br />
of its ancestors. For instance, the marginal probability of C4 2 9 indicates that <strong>based</strong><br />
on the current requirements and design, there is a probability of 0.70 (shown as 70% in<br />
Figure 10.6 by Netica) that the design is stable.<br />
181