Section: Research Program

Emergent mobile distributed systems

Emergent mobile distributed systems promise to provide solutions to the complexity of the current and future computing and networking environments as well as to the ever higher demand for ubiquitous mobile applications, in particular being a response to the volatile and evolving nature of both the former and the latter. Hence, such systems have gained growing interest in the research literature. Notably, research communities have been formed around self-adaptive systems and autonomic systems, for which various overlapping definitions exist  [72]. Self-adaptive systems are systems that are able to adapt themselves to uncertain execution environments, while autonomic systems have been defined as having one or more characteristics known as self-* properties, including self-configuring, self-healing, self-optimizing and self-protecting  [54]. Self-adaptive or autonomic systems typically include an adaptation loop comprising modeling, monitoring, analyzing, deciding and enactment processes. The adaptation loop provides feedback about changes in the system and its environment to the system itself, which adjusts itself in response. Current research on emergent distributed systems, including mobile ones, addresses all the dimensions of the adaptation loop  [31], [25], [61], [83].

In our previous work, we introduced the paradigm of emergent middleware, which enables networked systems with heterogeneous behaviors to coordinate through adequate interaction protocols that emerge in an automated way  [50], [28], [26]. A key point of that work is the combined study of the application- and middleware-layer behaviors, while current efforts in the literature tend to look only at one layer, either the application  [48] or the middleware  [19], [49], and take the other for granted (i.e., homogeneous, allowing direct coordination). Furthermore, the uncertainty of the computing and networking environments that is intrinsic to emergent mobile distributed systems  [41] calls for taking into account also the underlying network and computational resources in a cross-layer fashion. In another line of work, we studied cross-integration of heterogeneous interaction paradigms at the middleware layer (message passing versus event-based and data sharing), where we investigate functional and QoS semantics of paradigms across their interconnections  [43], [53]. Our focus there is to grasp the relation between individual and end-to-end semantics when bridging heterogeneous interaction protocols. In contrast, existing research efforts typically focus on emergent or evolving properties in homogeneous settings  [42]. Last but not least, integrating heterogeneous mobile distributed systems into emergent compositions raises the question of dependability. More specifically, the overall correctness of the composition with respect to the individual requirements of the constituent systems can be particularly hard to ensure due to their heterogeneity. Again, current approaches typically deal with homogeneous constraints for dependability  [39], [85], [40] with few exceptions  [38].

As evident from the above, there is considerable interest and intensive research on emergent mobile distributed systems, while at the same time there are key research questions that remain open despite initial relevant work, including ours, which are summarized in the following:

• How to effectively deal with the combined impact on emergent properties of the different functional layers of mobile distributed systems (e.g.,  [50], [28], [26], [69])?

• How to perceive and model emergent properties in space and in time across volatile compositions of heterogeneous mobile distributed systems (e.g.,  [43], [53])?

• How to produce dependable emergent mobile distributed systems, i.e., systems that correctly meet their requirements, despite uncertainty in their emergence and execution exacerbated by heterogeneity (e.g.,  [38])?