MARBLE™ (Modernization Approach for Recovering Business processes from LEgacy Systems) is a framework that focuses on the reverse engineering stage in the horseshoe modernization model to obtain business processes from legacy information systems. KDM is the core of MARBLE™, since it enables the representation and management of knowledge extracted by means of reverse engineering from all the different software artifacts of the legacy system in an integrated way. Then, that legacy knowledge is gradually transformed into business processes. For this purpose, MARBLE™ is divided into four abstraction levels with three transformations among them.
- Level L0. This level represents the legacy information system in the real world, and is the source system to recover underlying business processes.
- Level L1. This level represents several specific models, i.e., one model for each different software artifact involved in the archeology process like source code, database, user interface, and so on. These models are considered PSM models since they depict the software artifacts according to their specific technology platforms.
- Level L2. This level consists of a single PIM model that represents the integrated view of the set of PSM models in L1. The KDM metamodel is used for this purpose, since it makes it possible to model all the artifacts of the legacy system in an integrated and technological-independent manner. Firstly, we say that L2 is obtained in an integrated way because L2 works as a KDM repository that can be progressively populated with knowledge extracted from the different information systems of an organization. This is due to the fact that the structure package at the last KDM abstraction layer (c.f. Section 2.2) allows representing different systems, subsystems, components, architecture view, and so on. This is a key advantage because business processes are usually executed by multiple systems within an organization. Secondly, we say that L2 is represented in a technological-independent way due to the fact that KDM standard abstract (from the program element layer, the second layer of the KDM metamodel) all those details concerning the technological viewpoint (e.g. the program language). Thereby, the KDM repository at L2 can represent several legacy information systems even when their program languages or platforms are different.
- Level L3. This level depicts, at the end of the archeology process, the business processes recovered from the legacy system that is fully represented by a KDM model in L2. The business process model in L3 represents a CIM model of the system.
The three generic transformations between the four abstraction levels proposed in MARBLE™ are the following:
- Transformation L0-to-L1. This transformation obtains PSM models from each legacy software artifact. Classical reverse engineering techniques such as static analysis, dynamic analysis, program slicing and dicing, formal concept analysis, subsystem decomposition, and so on, could be used to extract the knowledge from any software artifact and build the PSM model related to it. These PSM models are represented according to specific metamodels. For example, a Java metamodel may be used to model the legacy source code, or an SQL metamodel to represent the database schema, etc.
- Transformation L1-to-L2. The transformation between levels L1 and L2 consists of a set of model transformations to obtain a PIM model based on the KDM metamodel. This PIM model is built from the PSM models from level L1. The L1-to-L2 transformation can be implemented by means of QVT. The transformation from the legacy information system (L0) to the KDM model (L2) is not direct due to the fact that, in many cases, the platform-specific knowledge in the intermediate level L1 might be used to infer more business knowledge. Thus, the semantic gap between the legacy system and its KDM model is reduced incrementally through L1.
- Transformation L2-to-L3. This transformation is based on a set of patterns. When a specific structure is detected in the KDM model from level L2, each pattern indicates what elements should be built and how they are interrelated in the business process model in L3. This is known as pattern matching and can be implemented in MARBLE™ by means of QVT relations, the declarative part of the QVT language. In addition, this last transformation can be assisted by business experts who know the organization. The external information provided by experts also serves as valuable knowledge in business process archeology. Experts can determine inconsistent or incoherent fragments in the preliminary business process models obtained after pattern matching; they can refactor the business process models, and incrementally fit the process models to the real behavior of the organization. Moreover, this transformation supports the representation of several subsystems at the same time. This transformation structures subsystems as different pools and lanes in the process model. Lanes and pools, which represent BPMN sub-modules, are related between them by means of certain data objects. These data objects are written by a sub-module and are read by another sub-module (e.g. data stored in a database by a subsystem and loaded by other subsystem; or a subsystem invoking a service of another one). Moreover, not all parts of business processes are executed by legacy information systems, i.e., there are a lot of manual business activities. Manual business activities can be also added by final business expert intervention. Since we are proposing an automated approach, the treatment of that kind of activities is beyond of the scope of this work. Other techniques address challenges like identifying human activities in the legacy system, activities outside the legacy system, identifying roles, and so on.
The main objective of MARBLE™ is to provide a first version of business process models that, compared with business process redesign by business experts from scratch, represents a more efficient solution and a good start point to achieve business process archeology. In addition, the business process redesign by experts from scratch might discard meaningful business knowledge that is only present in legacy information systems. Anyway, MARBLE™ supports the addition and integration of manual business activities together the discovered business processes. For this purpose, MARBLE™ defines a finite set of operations that business experts can carry out in the business processes at the end of the automate part of our method.
MARBLE™ is defined as a generic framework to rebuild business processes because it defines the organization of the four abstraction levels as well as the guidelines to perform the three transformations between them. However, this generic framework must be instantiated for each kind of program language on which the system is based, for each reverse engineering technique that is considered to extract the legacy knowledge, and so forth. The following subsections depict the three specific transformations that show how the business process archeology is carried out.