A Rationale-based Model for Architecture Design Reasoning

10.3. Reasoning about change impact with AREL
system, architects can investigate the possible causes (i.e. both architecture elements and
decisions) which contribute to the being of the design object. As such, considerations of
these causes could be made to ensure the design consistency.
10.3.5 Combining diagnostic and predictive reasoning
When an architecture is subject to impact analysis in a real-life project, it is often necessary
to understand the relationships between the design object which is undergoing modification
and other design objects which might be affected by such modification. The side-effects of
a modification may come from a few areas: (a) design objects which are directly affected
by the modification; (b) requirements, constraints or assumptions which are in conflict
with the modification; (c) other design objects that may be indirectly affected when the
requirements, constraints or assumptions are compromised.
Using BBN’s diagnostic and predictive reasoning, architects may use posterior probabilities
to guide the traversal of the AREL network to quantify likely changes within the
network. Based on the message processing example shown in Figure 10.6, we investigate
the changes that are likely to happen to the design of the system if we are to modify
C4 2 9.
C4 2 9 is responsible for sending acknowledgement messages to a bank that originates
a payment message. It is also responsible for receiving acknowledgement from a bank after
sending it payment messages. In either case, only when an acknowledgement is received
or sent would a payment message be considered legally binding. With the current design
of the system, the system has to reconcile what acknowledgements have not been sent
or received using a signalling mechanism. The design seems complicated and resource
intensive. We are to investigate what-if this mechanism is to change and which parts of
the system would be impacted.
Using the BBN network, we set the evidence and quantitatively analyse the likelihood
of change. Table 10.1 illustrates a sequence of reasoning steps to find out how setting
the evidences for C4 2 9 and other architecture elements could affect the validity of architecture
rationales and the volatility of architecture elements. This combined reasoning
process involves using both the BBN diagnostic reasoning and predictive reasoning. The
leftmost column in Table 10.1 shows the name of the BBN node (grouped by the AE nodes
in the upper set, and the AR nodes in the lower set). Each of the remaining columns shows
the change that is being investigated at each step, corresponding to the evidence being
entered into the BBN, by setting the appropriate AE volatility to 100%.
The following is a description of the reasoning steps that are illustrated in 10.1:
186