Have ReSpecT for LogOp

In this paper, we exploit two different coordination models, namely LOGOP [1] and ReSpecT [2], to discuss the design of a coordination infrastructure for distributed systems. LOGOP model introduces coordinated primitives extending those of LINDA towards usage on a multiplicity of tuple spaces, which can be dynamically associated by logical operators and then accessed altogether. These primitives are here given a mapping on top of the architecture of ReSpecT tuple centres, a coordination infrastructure suitable for distributed environments. A formal treatment of LOGOP primitives is presented that allows us to reason about the applicability of the resulting model in the context of distributed environments. I. BACKGROUND & MOTIVATION Complex systems of today are usually built out of several heterogeneous components (objects, processes, agents, . . . ) dipped into dynamic and distributed environments. Managing the space of the interaction among components is a matter of coordination [3]: starting from the pioneering work on LINDA [4], a number of different coordination models have been designed and proposed to handle the complexity of components interaction in system engineering [5]. Coordination models are typically conceived as providing a general solution to coordination issues: the LINDA model, for instance, has been experimented to face a wide class of different application problems. However, no coordination model is really general enough to effectively cover the whole spectrum of the different coordination problems that may arise from the engineering of complex systems. Every coordination model promotes a particular level of abstraction over systems, by providing a powerful but limited number of coordination abstractions through which the space of interaction is modeled and ruled. Consequently, several issues have to be addressed when choosing a coordination model that basically concern the expressive power of coordination models, and how do they compare to each other. In particular, given a coordination model, the first issue to be addressed is what class of coordination problems are actually addressed by the model, and how effectively. A measure of expressiveness for coordination models and languages is also required that allow designers to compare different models, and helps them choosing the right one for their particular application scenario (a first result in this direction is [6]). Furthermore, when complexity and applicability range of the coordination problem are large, it may be the case that more models have to be used together – how they could be combined, R. Menezes is with Department of Computer Science at Florida Institute of Technology, 150 West University Blvd., Melbourne, FL 32901, USA – email: [email protected] A. Omicini and M. Viroli are with DEIS, Università degli Studi di Bologna, via Rasi e Spinelli 176, 47023 Cesena (FC), Italy – email: {aomicini,mviroli}@deis.unibo.it merged, blended together in a more or less clean and effective way is also a relevant matter to be solved. According to the above considerations, in this paper we take two relevant examples of coordination models, namely LOGOP [1] and ReSpecT [2], that are both LINDA-based, and are seemingly meant to cover the same class of application problems. However, the difference in the approach and in the solutions they provide – namely, two opposite ways to extend the basic LINDA model – makes them complimentary models that can be used altogether at different levels of abstraction in the coordination of distributed systems. This paper is organized as follows: Section II introduces LOGOP in short, defining the basic coordination language and providing a first informal description of its semantics. After a brief discussion about the main features of ReSpecT in Section III, Section IV summarizes the differences between the two models, and describes a possible mapping of LOGOP upon ReSpecT. Finally, Section V introduces the issue of the semantics of coordination models, discussing the different models of the interaction space implicitly endorsed by different models of coordination, and also showing how a distributed and asynchronous environment could induce different interpretations of the intended effect of coordination primitives – which are not likely to be made unambiguous by any informal semantics.