Collaborative software engineering on large-scale models

Abstract

This work presents an approach for realizing Model-Driven software engineering in the distributed and multi-developers context. It particularly focuses on the scalability problems in a complex software project involving a large set of inter-connected models: (1) how to manipulate large data volume with limited computing resources, and (2) how to maintain consistency of inter-model links in a large model set, facing to concurrent model updates. As a solution, we propose the scalable copy-modifymerge mechanism, which allows each developer to copy only a model subset from the entire model set, to manipulate this subset locally, and to merge it back to the repository. This mechanism ensures the global consistency of the model set, particularly against dangling links. Our approach is generic: it is applicable to all model types (UML and Domain-Specific Models). Also, it offers interoperability with existing, heterogeneous CASE tools. Its prototype implementation in the ModelBus environment is now available on the Eclipse project "MDDi".

From subsets of model elements to submodels

“…Storing any modeling design effort. This facilitates incremental design, versioning, model sharing within design teams and finally traceability [2,36,47,49].…”

“…This process may produce models with "dangling edges" such as delta in the figure, when applied to m . Model differencing is currently used in model versioning [2,49], collaborative design [47] and more generally model management [36].…”

Ocarina : An Environment for AADL Models Analysis and Automatic Code Generation for High Integrity Applications

“…In addition, the use of UML and XML introduces a scalability issues when it comes to model applications with a large number of components [SBG08]. All these drawbacks restrict the use of COSMIC for DRE systems where correctness by construction is required.…”

On Differencing Object-Oriented Formal Specifications.

“…Collaboration [17] is necessary during the development of large-scale software systems where several developers work in parallel on different aspects of the same system. Often, this leads to the creation of different but related documents.…”

“…similarity scoring < 1) then there is a possibility that their contents (other than names) have been modified, new links have been attached to them or that some of their links have been removed. The algorithm addresses this by detecting the changed properties other than names and adding them to delta (line [13][14][15][16][17][18] 2 of type "derived_from", "aggregated_by" or "associated_with" connects it to a new class B and a link k 1 of the same type with an old class A is removed. A function updateMoving is used to perform the above verification; it accepts a delta containing a list of operations and a specification element E as parameters and returns an object containing the two elements representing the old and new link ends or a "null" object if no moving has taken place.…”