H. E. Bal and A. S. Tanenbaum. Distributed Programming with Shared Data. In IEEE Conference on Computer Languages, pages 82--91, IEEE, 1988.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....work, Kimura proposed the Transaction Network [19] a visual language in which the traditional places and transitions appearing in Petri nets have been replaced by databases and transactions, respectively. Recently, we became aware of several other groups working on similar kinds of languages [4, 24]. We use the term shared dataspace to refer to the general class of models and languages in which the principal means of communication is a common, content addressable data structure. Because the investigation of shared dataspace languages is still in its early stages, the body of knowledge ....

H. E. Bal and A. S. Tanenbaum. Distributed programming with shared data. In Proceedings of the 1988.

....by generalized path expressions, the irregular ones can not. Some coordinator features such as collective operations, however, are on a different i.e. more pragmatic level than path expressions. Coordinators share some concepts with object based distributed languages such as Emerald[3] and Orca[2]. See [1] for a general description of this kind of languages. While the automatic migration in these languages is similar it does not rely on algorithmic patterns. Note that coordinators are fully (i.e. recursively) migrated. These is akin to parameter passing in DCE[19] although a bit more ....

H. Bal and A. Tanenbaum. Distributed Programming with Shared Data. Proceedings of the IEEE CS 1988 International Conference on Computer Languages, 1988.

....tuple space. In principle, this programming model fits well to distributed computing. Certainly, it can be used for shared platforms as well. LINDA is abstract w.r.t both the process and the data model. There are some other approaches with similar properties, e.g. Emerald[BHJ 87] and Orca[BT88] although both are designed for distributed programming. In the following, we discuss these aspects in more detail in order to give a rationale for the model we have defined. 2.1 Process models A peculiarity of the LINDA programming model is the abstract process model. This model features ....

.... many ways (mutexes, semaphores, monitors) These primitives are present, among others, in the ANL Parmacs macros [LO87] and various thread packages, e.g. POSIX threads[P10] Logical ownership is supported in distributed programming within some approaches, too (such as Emerald[BHJ 87] and Orca[BT88] see [BST89] for an excellent overview) In the following, we present a programming model which allows programming on a more abstract level than the models described above without sacrificing performance. 3 The DIPP programming model We present a process and a data model which both try to ....

[Article contains additional citation context not shown here]

H. Bal and A. Tanenbaum. Distributed Programming with Shared Data. Proceedings of the IEEE CS 1988 International Conference on Computer Languages, 1988.

....[Black 85a] applications were programmed in EPL [Black 85b] an extension of Concurrent Euclid; the design of Emerald [Black 86a, Black 86b, Jul 88] was based on the Eden experience. The olus language [Wilkes 86] was developed for the Clouds distributed system [LeBlanc 85] The Orca language [Bal 88b] is being implemented to program applications in Amoeba [Tanenbaum 86, Mullender 87] Finally, the Comandos project is currently developing its own programming language, called Oscar [Comandos 88] Guide may be viewed as an exploratory implementation of some of the features being developed in ....

Bal H.E., Tanenbaum A.S., Distributed programming with shared data, Proc. IEEE CS 1988 Int. Conf. on Computer Languages, Miami, Fl., pp. 82-91 (oct. 1988)

....language techniques to provide programmers with further control over communication. The performance study shows that these features are important to achieve good performance in an update based implementation of entry consistency. Distributed object systems such as Emerald [9] Amber [5] Orca [2] and SAM [17] offer a shared address space of objects. However, on Emerald and Amber objects are not replicated. Orca [2] uses compile time analysis to decide whether to replicate shared objects in all the processors or not to replicate them. Write operations are broadcast to all the processors ....

....these features are important to achieve good performance in an update based implementation of entry consistency. Distributed object systems such as Emerald [9] Amber [5] Orca [2] and SAM [17] offer a shared address space of objects. However, on Emerald and Amber objects are not replicated. Orca [2] uses compile time analysis to decide whether to replicate shared objects in all the processors or not to replicate them. Write operations are broadcast to all the processors using a function shipping policy. DiSOM offers a more flexible programming model that provides finer grained control over ....

[Article contains additional citation context not shown here]

H. Bal and A. Tanenbaum. Distributed Programming with Shared Data. In IEEE Conference on Computer Languages, 1988.

....Another related paradigm developed by the distributed systems community is virtual shared memory, which provides the illusion of a single, shared, coherent address space for systems with physically distributed memories. Virtual shared memory models include page based [101] and object based systems [16, 46, 66, 89]. Page based virtual shared memory enforces consistency at the level of the memory page, typically one to four thousand bytes. Because such a coarse page granularity results in poor performance due to false sharing, object based systems provide consistency at the level of a single user defined ....

H. E. Bal and A. S. Tanenbaum, Distributed programming with shared data, in Proceedings of the International Conference on Computer Languages, October 1988, pp. 82--91.

....in DT i . A failure of such a site is crucial, because the CoK must wait until the site recovers. We show improvements of PR deep by providing alternative primary copies (see Section 4.2) Also other strategies can be designed. However, in contrast to the protocol used for a distributed Orca [1, 2] implementation, all sites below the current node in the DT i may fail without compromising the access of this CoK. A further advantage of communication objects is that as soon as a communication object i is written, the DT i can be resolved. All copies turn into read only ones that can be ....

.... n) PROCESS(LOCAL,timeout(TIME) pid T ) PROCESS(LOCAL, remote call(MDI 1 ,L 1 )IN(tid) pid 1 ) Delta Delta Delta PROCESS(LOCAL, remote call(MDIn,Ln )IN(tid) pidn) IF ALT WAIT( 1] COBJ READ(pid 1 ) SUCCEEDED AND Delta Delta Delta Delta Delta Delta AND COBJ READ(pidn) SUCCEEDED, [2] COBJ READ(pid 1 ) SUCCEEDED, Delta Delta Delta [n 1] COBJ READ(pidn ) SUCCEEDED, n 2] COBJ READ(pid T ) 1 THEN TRANS ABORT(tid) TRANS END gta(tid, TRUE) ENTRY remote call(ADDRESS:MDI;QUERY:q)IN(tid f ) USES(STAT OID:pid,tid;INT OID:cid;RES OID:rid) TRANS BEGIN(tid,tid f ) TRANS ....

[Article contains additional citation context not shown here]

H. Bal and S. Tanenbaum, "Distributed Programming with Shared Data," Computer Languages Journal, Vol. 16, No. 2, 1991.

....the number of cacheable objects is fixed and no general caching architecture is available. The second interesting feature of the Spring caching architecture is that caching is performed by a separate cacher domain. This is a different approach than that used in object oriented systems such as [4, 5, 6, 7] where each domain caches a copy of the objects that it is using. These systems are all designed around a particular object based programming system that supports distributed programming. Their models assume a single multi threaded address space where any operation outside the address space is ....

....is that caching is completely transparent to client applications. If an application program happens to use an object that supports caching, then the object will be automatically cached, without the application needing to be aware of it. Some previous implementations have provided this feature [5, 8] and others have not [4, 6, 7] The fourth interesting feature of the Spring caching architecture is that it does not rely on Global Unique Identifiers (GUIDs) for object equivalency. The ability to determine whether objects are equivalent is necessary so that the system can decide what objects ....

Bal, H. and A. Tanenbaum. "Distributed Programming with Shared Data," IEEE Conference on Computer Languages (October 1988): 82-91.

....hardware supported single address space systems can ease the task of parallel programming by eliminating separate address spaces and explicit communications. Examples of these systems are Amber [CAL 89] Clouds [RAK88] Dash [LLG 90] Ivy [LH89] Midway [BZ91] Munin [CBZ91, KCZ92] Orca [BT88] and Platinum [CF89] They preserve sequential semantics by enforcing a consistency protocol, which can be lazy or eager, based on invalidations or updates. Munin supports several such protocols, the choice between them for each individual shared variable is guided by access pattern annotations ....

Henri E. Bal and Andrew S. Tanenbaum. Distributed programming with shared data. In Proceedings of the IEEE CS 1988 International Conference on Computer Languages, pages 82--91, October 1988.

....variables must be declared in each process. In this example, the shared variable a is seen by both P 1 and P 2 , but private copies of c are held by each. Finally, we also note that these shared and private variables have visibility analogous to the shared and value declarations in Orca [BT88] 3.3. Barrier synchronization and merging private state While we have discussed how we can initially distribute information down from global state to private Tree structured state a, b, c c c b Shared Private p2 p1 ( var a, b, c; private c in p1 : var b . p2 ) ....

H. Bal and A. Tanenbaum, "Distributed programming with shared data," in Proceedings of the IEEE CS 1988 International Conference on Computer Languages, (Miami, Fla., Oct. 9-13), pp. 82--91, The Computer Society of the IEEE, New York, 1988.

....The CSP model [Hoa78] on which occam is based is a programming model rather than a language. Serializability. More advanced communication properties are required today. For example, if two messages are submitted in a specified sequence, they should arrive at their destinations in the same sequence [BT91]. The executions of simultaneous communications are serializable if each participant sees the same communication sequence. See the more general definition of serializability in [BHG87] Message delays cannot change the semantics of communicating programs that support serializability. The ....

....the more general definition of serializability in [BHG87] Message delays cannot change the semantics of communicating programs that support serializability. The languages at the serializability communication level give the programmer the impression that message delivery does not take time. Orca [BT91] uses the ideas of guarded commands [Dij75] and shared objects in an imperative language. Concurrent logic languages [Gre87, Sha89] like Concurrent Prolog, Flat Concurrent Prolog, Guarded Horn Clauses, and Parlog combine the CSP model with a logic programming language. The single assignment ....

[Article contains additional citation context not shown here]

H. Bal and S. Tanenbaum. Distributed programming with shared data. Computer Languages Journal, 16(2):129--146, 1991.

.... any programming language (e.g. Prolog plus Linda is known as Shared Prolog [6] C plus Linda is termed C Linda [10] Whereas message passing reflects the physical distribution, the shared variable philosophy can be supplied efficiently on shared as well as on distributed memory architectures [3, 1]. Shared variable communication provides a high level abstraction of the underlying architecture. This is quite a contrast to Occam, the language usually used with transputers. Occam shares no modifiable variables. Processes exchange information via channels, i.e. unidirectional data pipes. The ....

H. Bal and S. Tanenbaum. Distributed programming with shared data. Computer Languages, 16(2):129--146, 1991.

....issues such as replication and data translation, language developers can concentrate on the design of high level abstractions that simplify the programming of distributed applications. 3. 4 Orca Orca is a new language for distributed programming [4] It is based on the shared data object model [3], which has been designed to simplify the development of parallel applications in a distributed system. In the shared data object model, all shared data is encapsulated in passive objects which are instances of user defined abstract data types. Shared data may be accessed only by well defined ....

....an efficient reliable broadcast protocol [16] was developed for use by the run time system. The developers of Orca present convincing arguments that the shared data object model is superior to other mechanisms, including shared variables and message passing, for distributed process cooperation [3]. In addition, they contend that the shared data object model is a better approach to providing support for distributed data structures than is currently found in parallel languages such as Linda [15] Finally, they note that it is difficult to efficiently implement a language which supports the ....

Bal, H. and Tanenbaum, A., Distributed Programming with Shared Data, Proceedings 1988 International Conference on Computer Languages, IEEE, Oct., 1988, pp. 82-91.

....cannot deal with failures, which makes them unsuitable for heterogeneous environments. The approaches to exploit parallelism in Prolog implicitly are not of interest for coordination languages that need explicit language constructs to reason about the data and control flow of activities. Orca [3] [5] is a new object based programming language that uses the shared data communication paradigm. Orca provides atomic actions on distributed objects. There exist very detailed performance analyses of Orca implementations in the Internet, on a multi processor architecture and on the distributed Amoeba ....

.... (trip) travel grant( p[1] INDEP) process; 110 (pres2 grant.office; trip) travel grant( p[2] INDEP) process; 111 (pres3 grant.office; trip) travel grant( p[3] INDEP) process; 112 (LOCAL; p; request) wait for local; p[4] INDEP) process; 113 (LOCAL; p; trip) wait for grant; p[5]) INDEP) process; 114 (LOCAL; p; trip) wait for none; p[6] INDEP) process; 115 g 116 trans (comm int p[7] comm ToDoEntry request) wait for local( 117 f 118 if(request.result = YES) 119 request.task account.fund = LOCAL; 120 else abort; 121 g ( p[1] ABORT) signal, ....

[Article contains additional citation context not shown here]

H. Bal and S. Tanenbaum. Distributed programming with shared data. Computer Languages Journal, 16(2):129--146, 1991. 32 AUTHORS

....a SM simulation using the MP primitives. This can be done in two ways: The first method requires the construction of a compiler or preprocessor that will translate a shared memory reference to the appropriate sequence of MP commands that will satisfy the request by simulating hardware functions [5, 6, 7, 8]. This method is expected to gain wider acceptance in the near future, as compiler technology is rapidly evolving. In the future, compilers will be able to predict the communicational behaviour of the program better than they can today, and to act accordingly [9] The second method is the ....

Henri E. Bal and Andrew S. Tanenbaum. Distributed programming with shared data. Technical Report IR-149, Vrije Universiteit, March 1988.

....replication and data partitioning to make the distributed objects behave as if it were a single shared object. The most common example of a distributed object is a memory cell in a multiprocessor with caches, but many others have been designed and implemented for both distributed systems [Ell85, BT88, BHJ 87] and for multiprocessors [Luc87b, Dal86, CD90] To access a distributed object, these systems provide sophisticated run time support so that each object is given a single name; the run time system must determine what node in the distributed object should be used for a particular ....

H. Bal and A. Tannenbaum. Distributed programming with shared data. In International Conference on Computer Languages, 1988.

No context found.

H.E. Bal, A.S. Tanenbaum, "Distributed Programming with Shared Data," Proc. IEEE CS 1988 Int. Conf. on Computer Languages, pp. 82-91, Miami, FL (10-13 Oct 1988).

....memory on top of distributed memory hardware [30] Many other, more advanced DSMs have subsequently been built [6, 7, 11, 14,17, 23,26, 28, 36,37, 40,43, 44, 45] The DSM system studied in this paper is Orca. Orca s programming model was designed and implemented prior to most other DSM systems [5] and differs from other DSM models by being object based and language based. Orca encapsulates shared data in objects, and allows the programmer to define operations on objects, using abstract data types. In contrast, many other DSMs use physical entities such as pages or memory regions as the ....

....for these applications. A related design issue is how to keep all copies of a shared object coherent. Simply broadcasting all updates does not suffice, since update messages from different sources may arrive in different orders at the destination processors, resulting in inconsistencies [5]. To solve this problem, the Orca system uses totallyordered group communication [24] which guarantees that all group messages are received by all processors in the same total order. If a write operation is invoked on a replicated object, the invoker broadcasts the operation using totally ordered ....

H.E. Bal and A.S. Tanenbaum, "Distributed Programming with Shared Data," Proc. IEEE CS 1988 Int. Conf. on Computer Languages, Miami, FL, pp. 82-91 (Oct. 1988).

....efficiently on distributed systems. In our experience, however, implementing the algorithms efficiently using RPC is complicated. In this paper, we will look at an alternative model for distributed programming that supports logically shared data. This model, the shared data object model [Bal and Tanenbaum 1988], allows processes to share data without requiring physical shared memory. Also, we have designed a new programming language, Orca [Bal and Tanenbaum 1988; Bal et al. 1989] based on this model. The intent of this paper is to show that, for several applications, the model is both easy to use and ....

....we will look at an alternative model for distributed programming that supports logically shared data. This model, the shared data object model [Bal and Tanenbaum 1988] allows processes to share data without requiring physical shared memory. Also, we have designed a new programming language, Orca [Bal and Tanenbaum 1988; Bal et al. 1989] based on this model. The intent of this paper is to show that, for several applications, the model is both easy to use and efficient. We do so by describing an implementation of Orca on a loosely coupled system and reporting on our experiences in using this implementation for ....

[Article contains additional citation context not shown here]

Bal, H. E. and Tanenbaum, A. S., Distributed Programming with Shared Data, Proc. IEEE CS 1988 Int. Conf. on Computer Languages, pp. 82-91, Miami, Fl., Oct. 1988.

No context found.

H. E. Bal and A. S. Tanenbaum. Distributed Programming with Shared Data. In Proc. IEEE CS 1988 Int. Conf. on Computer Languages, pages 82--91, October 1988.

No context found.

Bal, H.E. and Tanenbaum, A.S., " Distributed Programming with Shared Data", Computer Languages, Vol. 16, no. 2, 1991, pp. 129-146.

No context found.

H. E. Bal and A. S. Tanenbaum. Distributed Programming with Shared Data. In IEEE Conference on Computer Languages, pages 82--91, IEEE, 1988.

No context found.

E.H. Bal and A.S. Tanenbaum, "Distributed programming with shared data", in Proc. of the ICCL, Miami, Florida (USA), IEEE, Computer Society Press, 1988, pp. 82-91.

No context found.

E.H. Bal and A.S. Tanenbaum. "Distributed programming with shared data," in Proc. of the ICCL, pp. 82-91, Miami, FL, IEEE, Computer Society Press, 1988.

No context found.

H. E. Bal and A. S. Tanenbaum. Distributed programming with shared data. In IEEE International conference on Computer Languages, pages 82-- 91, October 1988.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC