Keywords software engineering, metrics, software development
Start Date: 24-OCT-88 / Duration: 48 months
[ contact / participants ]
The aim of the MERMAID project was to improve support for estimators and project managers in the area of effort forecast, risk analysis and resource modelling. The approach is based on the definition of local software development environments, without prior assumptions of cost drivers or their relationship to effort/cost. This in turn is based on assessment of the most significant problems in the classical methods of cost estimation (such as COCOMO and FPA), and is a synthesis of DeMarco's concept of phase-based measurement and the Badly-Basile meta-model approach.
The approach was implemented by several increments of the MERMAID toolset. An initial Sun-based prototype was used to test the basic assumptions and hypotheses of MERMAID. The feedback from the user community was so positive that it was decided to proceed towards a commercially usable version, running on PCs with either MS-DOS and MS-Windows or OS/2. This version incorporates the improvements identified in the tests of the prototype.
A programme of research was carried out to improve the understanding of the nature of the relationship between observed development productivity and selected cost drivers. Data has been collected from software development projects both within the partners' organisations and externally from collaborating organisations. In parallel, extensive knowledge elicitation was carried out through interviews with experts in the field.
A generalised data model of the software development process and a comparative evaluation of existing cost estimation tools have been completed. Results of the knowledge elicitation have equally been produced. The initial prototype toolset was delivered in October 1990.
The testing of the MERMAID hypotheses was completed using this prototype. The underlying assumptions were proven to be generally correct. The research results have been translated into product specifications for the final, commercially-oriented MERMAID toolset.
The PC version of the MERMAID estimation toolset was produced in an initial version in October 1991, and in its improved version in October 1993. Both run on MS-DOS/MS-Windows and OS/2.
A system for the collection of the project data (DCSS: Data Collection and Storage System) has been implemented on PC and has been used to analyze user project data.
In the area of risk analysis, a prototype has been produced which formed the basis for a refined PC-based version that was delivered in October 1993.
On the subject of resource modelling, two models were developed and refined, but the work in this area has not been mature enough to produce usable tools.
All partners (industrial, research and university partners alike) are exploiting their improved understanding of cost estimation in their consultancy services.
Cap Volmac has developed the PC-based estimation toolset in a commercial product, called ESTEEM, which has been commercially available since the second half of 1992.
Data Management intends to use the tools internally for the development of banking software.
NCC has developed the DCSS into a commercial product, called M-BASE, which has been commercially available since the middle of 1992.
VTT has developed the risk analysis tool into a commercially usable version.
The academic partners, City University and University College Cork will use the research result to expand their consultancy and education programmes.
PO Box 2575
NL - 3500 GN Utrecht
tel: + 31 / 30 526526
fax: +31 / 30 543244
CAP VOLMAC - NL - C
CITY UNIVERSITY LONDON - UK - P
NCC - NATIONAL COMPUTING CENTRE - UK - P
VTT - TECHNICAL RESEARCH CENTER
OF FINLAND - FL - P
UNIVERSITY COLLEGE CORK - IRL - P
DATA MANAGEMENT SPA - I - P
MERMAID - 2046, December 1993
please address enquiries to the ESPRIT Information Desk
html version of synopsis by Nick Cook