Page 204 -

DOMAIN-SPECIFIC & IMPLEMENTATION-INDEPENDENT SOFTWARE ARCHITECTURES 189
than 5, it may significantly alter the value of another metric, such as “DRAC Size,” that it is no
longer in an acceptable range. Thus, such tight constraints may mandate continued refinement
without ever being able to achieve some equilibrium state for the architecture.
RELATED WORK

This section highlights related work in software architecture evaluation methods and object-oriented
analysis and design as related to software architecture derivation and evaluation.

Software Architecture Evaluation

Approaches for software architecture evaluation cover a broad spectrum. On the one hand, evalua-
tion approaches can be classified based on the qualities they emphasize—from methods that focus
on particular qualities to broad-brush methods that provide a general evaluation framework. On
the other hand, approaches can be categorized based on the evaluation technique employed, from
static methods such as scenario analysis or metrics-based assessment to dynamic methods such
as simulation. While dynamic methods often provide an accurate measure of system performance
and other runtime qualities, such analysis is not relevant to the RARE DRA derivation and evalu-
ation process and will not be addressed in detail. RARE evaluation focuses on the measuring
characteristics of the architectural structure that results from defining DRACs and allocating
DM functionality and data to those DRACs. Thus, static approaches are more applicable to the
RARE process.
Much of the software architecture research has progressed as a result of individual communities
focusing on specific software system quality issues such as maintainability (Bengtsson and Bosch,
1999; Briand, Morasca, and Basili, 1993), comprehensibility (Briand, Morasca, and Basili, 1993),
reliability (Abd-Allah, 1997; Wang, Wu, and Chen, 1999), performance (Abd-Allah, 1997), inte-
grability (Abowd et al., 1993), reusability (Zhao, 2000), and flexibility (Lassing, Rijsenbrij, and
Van Vliet, 1999). RARE DRA derivation is concerned with structure, primarily the identification
of classes and the allocation of data and functionality. Consequently, these specific evaluation
methods must be applicable during the allocation activity to be relevant to RARE DRA evalua-
tion. For example, the recommendation by Briand, Morasca, and Basili (1993) for low average
coupling and high average cohesion to encourage maintainability and comprehensibility can be
directly applied to DRA data and service allocation—in fact, coupling/cohesion can be consid-
ered the strongest influence in DRA derivation. However, the reliability approaches suggested by
Abd-Allah (1997) and Wang, Wu, and Chen (1999) cannot be applied in whole since they rely
heavily on the semantics of architectural styles, some of which imply implementation properties
not addressed in the systems engineering process activities SEPA DRA.
While it is important for research efforts to focus on methods for evaluating individual quali-
ties, real-world software systems require a balance of different software qualities (Bengtsson and
Bosch, 1999; Bosch and Molin, 1999). Two approaches have been proposed for static evaluation
of architectures that are not specific to a particular quality: scenario-based methods and metrics-
based methods.
The most prevalent overall architecture evaluation approaches are rooted in scenario-based
methods such as the Software Architecture Analysis Method (SAAM) (Kazman et al., 1994) and
the Architectural Tradeoff Analysis Method (ATAM) (Kazman et al., 1998; Lougee, 2005). The
success of scenario-based approaches has led to the development of a number of related evalua-
tion methods (Asundi, Kazman, and Klein, 2001; Bengtsson, 2002; Dobrica and Niemela, 2002;