Notes and Further Reading

As mentioned earlier, Filter Pipeline is the basic architecture used in the UNIX operating system to combine programs. The architecture is also used extensively in multimedia applications to process streams of video and audio information. An example of this sort of program can be found in Kleiman, Shah and Smaalders (1996). A discussion of the properties of more general pipe and filter networks may be found in Shaw and Garlan’s book on Software Architecture (1996).

The Supervisor – Worker architecture appears in many guises in books and papers on concurrent programming. Andrews (1991) calls the architecture “replicated worker”, Burns and Davies (1993) call it “process farm”, while Carriero and Gelernter (1989b) characterize the form of parallelism supported by the architecture as “agenda parallelism”. A paper by Cheung and Magee (1991) presents a simple way of assessing the likely performance of a sequential algorithm when parallelized using the Supervisor – Worker architecture. The paper also describes a technique for making the architecture fault-tolerant with respect to worker failure. The Supervisor – Worker architecture has been used extensively in exploiting the computational power of clusters of workstations.

A large literature exists on Linda (Gelernter, 1985; Carriero and Gelernter, 1989a, 1989b) and its derivatives. The proceedings of the “Coordination” series of conferences describe some of the current work on tuple-space-based models for concurrent programming, starting from the early conferences (Ciancarini and Hankin, 1996; Garlan and Le Metayer, 1997) and including more recent events (Nicola, Ferari and Meredith, 2004; Jacquet and Picco, 2005). The Linda tuple space paradigm clearly influenced the work on JavaSpaces™ (Freeman, Hupfer and Arnold, 1999).

Event-based architectures have been used to connect tools in software development environments (Reiss, 1990). As discussed in the chapter, windowing environments are usually event-based, as Smalltalk (Goldberg and Robson, 1983). In a distributed context, event processing forms an important part of network management systems (Sloman, 1994). Shaw and Garlan (1996) discuss some of the general properties of event-based systems.