Nowadays, organizations are more and more faced with the need of modernizing their legacy systems to newer platforms.
Particularly, current legacy systems are mostly written in COBOL, whereas the target paradigm for migrating these systems is commonly SOA (Service-Oriented Architecture) due to its widely-recognized beneﬁts in terms inter operability and reusability.
A question that arises is, however, which is the best way to painlessly moving from a legacy system to exploit the advantages of SOA, since migration is in general an arduous endeavor for any organization.
In this paper, through a real-world case study involving the migration of an old CICS/COBOL system to SOA and .NET, we have shown that the traditional “fast and cheap” approach to migration –i.e. direct migration– produced a not-so-clear SOA frontier. Therefore, the resulting services were hard to reason about and consume by client application developers.
Alternatively, through an indirect migration approach, a better SOA frontier in terms of service quality was obtained, at the expense of much higher development costs.
The common ground for comparison was basically an established catalog of WSDL anti-patterns that are known to hinder service under standability and discoverability.
All in all, an interesting ﬁnding from this experience is that there is a relationship between the approach to migration to SOA used and the number of anti-patterns found in the resulting SOA frontiers.
Motivated by the high costs of indirectly migrating that system by hand, we also proposed a semi-automatic tool to help development teams in migrating COBOL systems to SOA.
Our tool comprises a number of heuristics that detect bad design and implementation practices in legacy systems, which in turn serve as a mean to propose early code refactorings so that the ﬁnal SOA frontier is free from as much WSDL antipatterns occurrences as possible.
To evaluate the approach,we used the CICS/COBOL system mentioned above. In the end, results were encouraging, since migration costs were dramatically reduced and service quality was very acceptable and close to that of indirect migration.