Monday, May 4, 2009
Yuanyuan Song
Chair: John Knight
Advisor: Kevin Sullivan, John Stankovic, Kim Hazelwood, David Garlan
OLSSON 236D, 15:30:00
A Ph.D. Proposal
DRAPES: Modeling and Analysis of Structurally Complex Design Spaces
ABSTRACT
One of the most important properties of a software design is adaptive capacity. Speaking informally, we view adaptive capacity as the extent to which a design can be changed in ways that are likely to create economic value net of the costs of change. Key challenges in practice include determining the value of adaptive capacity, determining the cost of adaptive capacity, setting and representing requirements and specifications for adaptive capacity, and verifying that designs meet such requirements and specifications. A specific research problem that my research group has been addressing is that we lack adequate ways of representing designs in ways that support modeling and analysis of their adaptive capacity. We are addressing this problem by developing, evaluating and refining theoretical constructs that connect decision-theoretic models of design structure to economic models of the value of flexibility in design. Previous work led to a finite state approach to modeling designs spaces, the relationship between modularity and adaptive capacity, and the real options value of adaptive capacity based on the models of Baldwin and Clark. The specific technical problem that I propose to address is that work to date does not yet adequately support the modeling of structurally heterogeneous design spaces: spaces in which different decisions at an abstract level yield structurally dissimilar spaces within which subsequent decisions must be made. Architectural decisions are an important class of such decisions. I propose to develop and evaluate a finite-state model of design space structure that addresses this problem. I call the proposed models discrete ramified phase spaces(DRAPES). DRAPES models are based on finite state modeling of design spaces and the well defined mathematical concept of bundles. DRAPES model hierarchical decisions as refinements of individual decisions in solutions to abstract problems. DRAPES models are able to capture decisions with structural consequences that otherwise hard to model. In this proposal, I present the DRAPES approach; I summarize results to date; and I describe my plans for completing the work required for the Ph.D. degree. In particular, I describe the mathematical model of DRAPES based on constraint bundles, the representation of DRAPES as structured sets of parameterized Alloy modules, the derivation from such models of finite-state models of adaptation dynamics, and the connection of these dynamic models to net options value models of economic value based on the earlier work of Cai and Sullivan. Among my results to date is a prototype tool supporting the creation and analysis of DRAPES models.