TY - JOUR TI - Making SOA-based systems coherent and trustworthy DO - https://doi.org/doi:10.7282/T30C4TQR PY - 2012 AB - Under Service Oriented Architecture (SOA), a software system - such as one that supports an enterprise-consists of multiple heterogeneous servers. These servers may be distributed over the internet, and may be managed under different administrative domains. SOA has become hugely popular, particularly as the architecture of large and complex distributed systems such as enterprise systems, grids, virtual enterprises, and supply chains. Unfortunately, this architecture as it is currently defined and being used, suffers from serious problems as outlined below. First, there is no means to ensure that a fragmented and open system like a SOA-based system satisfies desired global constraints, or can establish any regularity over the system. Second, service providers can make their own commitments to clients. However, the SOA methodology provides no guarantee to clients that the commitment made by a given server will be satisfied. Note that although service providers should have the freedom to specify their own commitments, they must conform to the global constraints imposed at large, and such conformity must be enforced. Third, when using a composite service, clients indirectly communicate with services via an orchestrator. Such indirect interaction may cause concerns for both clients and services, but no formal mechanism has been designed to address those concerns. Finally, it is not possible to ensure that coordination between disparate servers—called ”choreography” under SOA— is carried out safely and correctly. The overarching goal of my dissertation is to design a regulatory mechanism to address all these problems in a scalable manner. The mechanism is an extension of Law Governed Interaction (LGI) - a decentralized coordination and control mechanism for distributed systems. We call this mechanism ”LG-SOA”, for Law-Governed-SOA, which enables high expressive power, efficient enforcement, as well as good scalability. We will also present how LG-SOA can be applied to legacy systems. Case studies in the context of enterprise systems demonstrate the flexibility and applicability of this mechanism. Experiments show the overhead introduced by LG-SOA is relatively small, especially in the context of geographically distributed systems like SOA-based systems. In sum, LGI-SOA is effective and versatile in making SOA-based systems more coherent and trustworthy. KW - Computer Science KW - Service-oriented architecture (Computer science) LA - eng ER -