ON THE STUDY OF DYNAMIC AND ADAPTIVE DEPENDABLE DISTRIBUTED SYSTEMS

Due to the usage of MANETs and some kinds of collaborative applications (P2P), current distributed systems are becoming increasingly dynamic; i.e., it is difficult to manage membership information and to forecast the accessibility of each system node. Moreover, dependable applications for static distributed systems also need to provide good adaptability levels (to different request arrival rates, usage patterns, classes of requests,...) and good scalability; a case to study is the cloud computing paradigm. Development of dependable applications in dynamic and adaptive systems is not trivial, since both dynamism and adaptability may compromise algorithm liveness or may complicate the design of such algorithms, specially those best suited for static systems. Strategies for building adaptable and scalable dependable services (based on “cloud systems”) will be surveyed and improved. Moreover, an efficient support for dependable applications in dynamic systems will be provided, combining three different approaches: relaxed consistency models, interconnection protocols (for supporting both consistency and multicasting) and reconciliation strategies. Last but not least, also the usage and support for integrity constraints in replicated systems will be analyzed and improved for dynamic systems.