Deployment Model-Driven Architecture (D-MDA)

Motivation

The motivation of this project is to improve the process of developing / deploying individual IT solution architectures for enterprise customers of CA. The IT solution architecture is defined by a set of units, e.g. clients and servers, and the interconnection of these units. These units are equipped with proper applications in order to fulfill the customers business demands. Further, these IT solution architectures have to fit certain requirements given by customer demands. The current problem is, there is no common technique for describing IT solution architectures, neither there exists a common technique for analyzing the quality of these IT solution architectures, which would enable the comparison of different alternatives. Currently, the design of IT solution architectures is provided textual or by some kind of visual diagrams without any definition of a common semantic. Reviewing or analyzing these IT architectures is different for each solution and therefore related to increasing efforts in order to discuss / reason design decisions. Furthermore, this issue results in the lack of reuse of previously designed IT solution architectures or certain parts of these solutions. Another shortcoming is that it is a tedious task to predict if a solution fulfills its customer demands and further to predict what the impacts of certain architectural modifications are.

Challenge

To overcome the above-mentioned issues, several challenges arise which have to be treated. The first challenge is to provide a modeling technique, which captures all elements of interest of IT solution architectures. A model driven approach allows reuse previous designs or parts of them. Further, the model can provide a common definition of semantic and syntax, and a common understanding of IT solution architectures. Providing a model driven approach also enables automated model analysis, which is in this case, concerned to the fulfillment of certain requirements of interest. The outcome of these analyses can be utilized to study the impact of certain modifications. The analysis, which estimates the quality of IT solution architectures, requires evaluating several metrics first. Therefore, it has to be studied which metrics are necessary and which informations are required in order to evaluate these metrics properly. Metrics are estimated by means of evaluation methods, which can vary from simple analytic functions to complex simulations. The whole approach should benefit from various existing evaluation methods, which are developed in the past [1]. This requires the definition of a common interface, which allows the integration of various evaluation methods. For distinctive reasoning about quality, a framework is required which allows the definition of functional and non-functional requirements. Such a formal framework is essential because it permits to relate requirement issues directly to quality terms. Thus, the quality of IT solution architectures is directly related to the fulfillment of the requirements, which are of interest.

Benefits

The benefits of an approach, which attends the mentioned challenges, are significant. For example, it would improve the "Time-to-Value" because it permits the reuse of existing IT solution architectures and therefore provides a common access to knowledge of CATs and R&Ds. Furthermore, it permits to define best practices in a common formalism as also reuse measurements, which are related to certain elements of the model. In combination with the reusability of evaluation methods, IT solution architecture modifications can be directly assessed in order to study the impact on its quality term. Because of its common metrics, several IT solution architectures, which addressed to be comparable, can be compared in order to reduce testing efforts in a later life cycle. The efforts and the related costs for testing are reduced because the number of possible alternative IT solution architectures can be reduced by creating a pre-selection containing only these, which are suited well to the customer's requirements. Another benefit would be that efficient Enterprise IT Management (EITMTM)¹ installation configurations can be automatically derived.

Project

The entire project is a collaboration between the Haifa University in Israel and the department of system analysis and modeling at the HPI in Potsdam. In this collaboration the current practices of CAs EITMTM are being analyzed in order to elicitate crucial requirements of different stakeholders. Further, a model-driven approach is going to be developed, which allows proper modeling of IT solution architectures, and which has capabilities of capturing measurements that are required for evaluating estimated metrics, used to assess the fulfillment of certain related requirements. The entire project is set up to three years.

¹ See www.ca.com/eitm/

[1] Simonetta Balsamo, Antinisca Di Marco, Paola Inverardi and Marta Simeoni. Model-Based Performance Prediction in Software Development: A Survey. In IEEE Transactions on Software Engineering, Vol. 30(5):295-310, IEEE Computer Society, Los Alamitos, CA, USA, 2004.