Electronic International Standard Serial Number (EISSN)
Next generation distributed real-time systems will be complex high-performance environments containing applications with a flexible structure, integrating a large number of nodes of heterogeneous nature characterized by multiple and decoupled software units scattered all over the distributed environment; they are expected to offer data-intensive capabilities through merging the Processing power of large numbers of nodes. These systems will have increased dynamic behavior by suffering frequent reconfigurations or state transitions resulting, among others, from the changing nature of the processed data. Handling the dynamics of these systems in real-time is a complex problem that requires to impose some bounds to the structure of the system to really achieve timely response not only during normal operation but also in the event of reconfigurations. In this paper, we present an approach to achieve real-time reconfiguration in distributed real-time service-based systems modeled as graphs. A reconfiguration requires to search for a new schedulable/valid solution or state from a complete system graph that contains all tentative solutions; each of these solutions will have undergo a schedulability analysis to determine if it is a valid solution; if the system graph is too complex, the overall time required for the schedulability check can be exponential with respect to the size of services and service implementations; this may lead to an unbounded reconfiguration time. In this paper, we present an approach to reduce the complexity of the system graphs so that a summarizing one that contains valid solutions is analyzed and not the complete system graph. We have implemented this mechanism inside the iLAND service reconfiguration and composition components to validate the proposed concepts and ideas; the reduction of the space of solutions with the presented approach is very high, which dramatically decreases the computation time of the reconfiguration process.