ACM Computing Surveys 28A(4), December 1996, http://www.acm.org/surveys/1996/MisraDiscipline/. Copyright © 1996 by the Association for Computing Machinery, Inc. See the permissions statement below.
Abstract: My current research is based on two observations: (1) the applications that will be implemented on networks of processors in the future will be significantly more ambitious than the current applications (which are mostly involved with transmissions of digital data and images), and (2) many of the programming concepts developed for databases, object-oriented programming and designs of reactive systems can be unified into a concise model of distributed programs that can serve as the foundation for designing these future applications.
Categories and Subject Descriptors:
General Terms: Distributed Programming, Multiprogramming, Reactive Systems
Additional Key Words and Phrases:
We are currently working on a project, called Seuss. The research proposed in Seuss is based on two observations: (1) the applications that will be implemented on networks of processors in the future will be significantly more ambitious than the current applications (which are mostly involved with transmissions of digital data and images), and (2) many of the programming concepts developed for databases, object-oriented programming and designs of reactive systems can be unified into a concise model of distributed programs that can serve as the foundation for designing these future applications.
Research in multiprogramming has, traditionally, attempted to reconcile two apparently contradictory goals: (1) it should be possible to understand a module (e.g., a process or a data object) in isolation, without considerations of interference by the other modules, and (2) it should be possible to implement concurrent threads at a fine level of granularity so that no process is ever locked out of accessing common data for long periods of time. The goals are in conflict because fine granularity, in general, implies considerable interference. The earliest multiprograms (see, for instance, the solution to the mutual exclusion problem in [Dijkstra 1965]) were trivially small and impossibly difficult to understand, because the behaviors of the individual processes could not be understood in isolation, and all possible interactions among the processes had to be analyzed explicitly. Since then, much effort has gone into limiting or even eliminating interference among processes by employing a variety of synchronization mechanisms: locks or semaphores, critical regions, monitors and message communications.
Constraining the programming model to a specific protocol (binary semaphores or message communication over bounded channels, for instance) will prove to be short-sighted in designing complex applications. More general mechanisms for interactions among modules, that include these specific protocols, are required. Further, for the distributed applications of the future, it is essential to devise a model in which the distinction between computation and communication is removed; in particular, the methods for designing and reasoning about the interfaces should be no different from those employed for the computations at the nodes of the network.
Seuss fosters a discipline of programming that makes it possible to understand a program execution as a single thread of control, yet it permits program implementation through multiple threads. As a consequence, it is possible to reason about the properties of a program from its single execution thread, whereas an implementation on a specific platform (e.g., shared memory or message communicating system) may exploit the inherent concurrency appropriately. A central theorem establishes that multiple execution threads implement single execution threads, i.e., any property proven for the latter is a property of the former as well.
A major point of departure in Seuss is that there is no built-in concurrency and no commitment to either shared memory or message-passing style of implementation. No specific communication or synchronization mechanism, except the procedure call, is built into the model. In particular, the notions of input/output and their complementary nature in rendezvous-based communication ([Hoare 1984] , [Milner 1989]) are outside this model. There is no distinction between computation and communication; process specifications and interface specifications are not distinguished. Consequently, we do not have many of the traditional multiprogramming concepts such as, processes, locking, rendezvous, waiting, interference and deadlock, as basic concepts in our model. Yet, typical multiprograms employing message passing over bounded or unbounded channels can be encoded in Seuss by declaring the processes and channels as the components of a program; similarly, shared memory multiprograms can be encoded by having processes and memories as components. Seuss permits a mixture of either style of programming, and a variety of different interaction mechanisms -- semaphore, critical region, 4-phase handshake, etc. -- can be encoded as components.
Seuss proposes a complete disentanglement of the sequential and concurrent aspects of programming. We expect large sections of code to be written, understood and reasoned-about as sequential programs. We view multiprogramming as a way to orchestrate the executions of these sequential programs, by specifying the conditions under which each program is to be executed. Typically, several sequential programs will execute simultaneously; yet, we can guarantee that their executions are non-interfering, and hence, each program may be regarded as atomic. We propose an efficient implementation scheme that can, using user directives, interleave the individual sequential programs with fine granularity without causing any interference.
A few chapters from the document A Discipline of Multiprogramming are available online.
Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers, or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from Publications Dept, ACM Inc., fax +1 (212) 869-0481, or firstname.lastname@example.org.