A. Romanovsky, B. Randell, R. Stroud, J. Xu and A. Zorzo, "Implementation of Blocking Coordinated Atomic Actions Based on Forward Error Recovery," in Journal of Systems Architecture, vol. 43, pp.687-699, 1997.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....be needed, would have to provide more sophisticated features: aborting any number of nested actions, executing the abortion handlers, signaling exceptions from the nested actions, and involving all participants of the nested actions in the abortion activity, etc. Pre emptive CA action schemes [Romanovsky et al. 1997] provide these features, and can be implemented by the decentralised protocol described in [Xu et al. 1998] The CA action framework allows us to express various time related outcomes of a CA action. If error recovery after a real time exception has been successful within the action, then its ....

A. Romanovsky, B. Randell, R. Stroud, J. Xu and A. Zorzo, "Implementation of Blocking Coordinated Atomic Actions Based on Forward Error Recovery," in Journal of Systems Architecture, vol. 43, pp.687-699, 1997.

....In Section 4, we describe the specification and verification methodology for our system. Finally, we conclude with notes on future work in Section 5. 1. 1 Related Work Much work has been done on checkpointing and logging algorithms (e.g. 3, 6, 17, 35, 40, 47] backward and forward recovery [15, 16, 37, 38], and compositional specification and verification [2, 7, 10, 12, 21, 24, 25, 34, 36, 41, 46] In this section, we list some of the more closely related work. Neves and Fuchs [31] consider the problem of recovery in mobile environments. The solutions they propose are also applicable to the ....

A. Romanovsky, B. Randell, R. Stroud, J. Xu, and A. Zorzo. Implementation of blocking coordinated atomic actions based on forward error recovery. Journal of Systems Architecture, 43(10):687--99, September 1997. 20 James, Singh

....can give a partial view of the state of the multiset; 4] A user may exit the system provided he has entered it. State and internal behavior: 5] The integers entered by the users are stored in a multiset; 6] The application realizes the sum of all the integers entered by all the users; [7] The sum is performed by chemical reactions according to the Gamma paradigm; 8] A chemical reaction removes two integers from the multiset, adds them up, and inserts the sum into the multiset; 9] There is only one type of chemical reaction, but several of them can occur simultaneously and ....

....a client server application implemented with the Java programming language. In order to experiment our methodology with other design constraints, we have chosen to develop another implementation of the DSGamma system using Coordinated Atomic Actions (CAAs) 5] CAAs are multi threaded transactions [9, 7]: 1) threads inside a CAA may access global objects; if threads belonging to two or more CAAs access the same global objects, the CAA paradigm ensures that these objects are accessed according to the ACID properties; 2) threads inside a given CAA may communicate with each other according to the ....

A. Romanovsky, B. Randell, R. Stroud, J. Xu, and A.Zorzo. Implementation of Blocking Coordinated Atomic Actions Based on Forward Error Recovery. Journal of System Architecture - Special Issue on Dependable Systems, July 1997.

....2 object role 3 object Figure 4: Distribution of CA action components We have chosen to treat individual roles within an action as the units of distribution because this allows them to be executed in the locations at which information is produced or consumed. Although several approaches (e.g. [6]) view CA actions as packages (modules, objects, etc. we do not adhere to this approach because these units cannot be split into parts and distributed. The 4 general idea of attaching exception handling to roles suits role distribution very well, in spite of the fact that these handlers are ....

A. Romanovsky, B. Randell, R. Stroud, J. Xu, and A. Zorzo, "Implementation of Blocking Coordinated Atomic Actions Based on Forward Error Recovery". Journal of System Architecture (to be published in July/97). 5

....we identify the choices for designing an ideal exception handling model suitable for developing this kind of systems. The criteria and the proposed exception handling model have been developed using our extensive experience in designing both dependable object oriented systems [Ferreira, 1999] [Romanovsky et al. 1997] [Rubira, 1994b] Xu et al. 1995b] Xu et al. 1995a] Xu et al. 1998] Xu et al. 1999] and the exception mechanisms suitable for this kind of systems [Garcia et al. 1999] Garcia, 2000] Garcia et al. 2000] Romanovsky, 1997] Romanovsky et al. 1996] Xu et al. 1995a] Zorzo et al. ....

.... both dependable object oriented systems [Ferreira, 1999] Romanovsky et al. 1997] Rubira, 1994b] Xu et al. 1995b] Xu et al. 1995a] Xu et al. 1998] Xu et al. 1999] and the exception mechanisms suitable for this kind of systems [Garcia et al. 1999] Garcia, 2000] Garcia et al. 2000] [Romanovsky, 1997] [Romanovsky et al. 1996] Xu et al. 1995a] Zorzo et al. 1999] Xu et al. 2000] 6.1 Quality Requirements Q1. Readability. One of the main reasons for using exception mechanisms is improving the program readability [Burns and Wellings, 1996] Mitchell et al. The importance of readability ....

[Article contains additional citation context not shown here]

Romanovsky, A., Randell, B., Stroud, R., Xu, J., Zorzo, A., Implementation of Blocking Coordinated Atomic Actions Based on Forward Error Recovery, Journal of Systems Architecture 43, 687-699 (1997).

....the creation of the instances of these actions. This design is very general, so we assume that there is a set of hosts on which the instances of CA actions can be instantiated, and that the scheduler knows their location. We use a centralised CA action scheme, so each action has an action manager [13]. As we have mentioned before, CA actions provide a framework for dealing with exceptions that happen during the execution of the system. In our DSGamma system, when a fault happens inside a GammaAction, a predefined exception ReactionException is raised in the thread executing a role in the ....

A. Romanovsky, B. Randell, R. Stroud, J. Xu, and A. Zorzo. "Implementation of Blocking Coordinated Atomic Actions Based on Forward Error Recovery". Journal of System Architecture - Special Issue on Dependable Parallel Computing Systems, 43(10), pp. 687-99, 1997.

....of the instances of these actions. This design is very general, so we assume that there is a set of hosts on which the instances of the CA actions can be instantiated and that the scheduler knows their location. We will use a centralized CA action scheme, so each action has an action manager [3]. 4.1.5 InsertNumberAction and FinishAction Apart from GammaAction, we have introduced two other CA actions: InsertNumberAction and FinishAction. These actions are executed by the TGetNumbers thread. When a new number is entered, TGetNumbers executes a role inside InsertNumberAction. When the ....

A. Romanovsky, B. Randell, R. Stroud, J. Xu, and A. Zorzo. "Implementation of Blocking Coordinated Atomic Actions Based on Forward Error Recovery". In Journal of System Architecture - Special Issue on Dependable Systems. July/97.

....shows how an instance of the GammaAction CA action would be initialised: We have chosen to treat individual roles within an action as units of distribution because this allows them to be executed in the locations at which information is produced or consumed. Although several other approaches (e.g. [11] [12] view CA actions as packages (modules, objects, etc. such packages cannot be split into parts and distributed. The general idea of attaching exception handling to roles suits role distribution very well, in spite of the fact that these handlers are controlled by the application independent ....

A. Romanovsky, B. Randell, R. Stroud, J. Xu, and A. Zorzo, "Implementation of Blocking Coordinated Atomic Actions Based on Forward Error Recovery". In Journal of Systems Architecture, 43, pp.687-699, 1997.

....The main rules of nesting are simple: sibling actions cannot overlap; if a process takes part in an action, it has to take part in the father action; the action is over only if all of its nested actions are. In the last two years several CA action schemes have been designed using Ada 95 and Java [Romanovsky 97a, Romanovsky 97b] They are presented as sets of reusable components (classes, generic procedures, packages, objects) and of templates and conventions which should be followed by programmers. These schemes differ in the ways of packaging actions and participants (e.g. actions can be associated ....

A. Romanovsky, B. Randell, R. Stroud, J. Xu, and A. Zorzo. "Implementation of Blocking Coordinated Atomic Actions Based on Forward Error Recovery". Journal of System Architecture 43(10).September, 1997. pp. 687-699.

....a simple example where numbers from distributed multisets are summed and the result is stored in a multiset. Our distributed GAMMA system is composed of a set of participants (located on different hosts) a scheduler (located on a separate computer) and a set of CA actions, called GammaActions [Romanovsky Zorzo 1997]. The design has two levels. The first level is concerned with information exchange between computers (participants and a scheduler) This is the level on which the execution of the GammaActions is scheduled (or the actions are glued together) At the second level of the design, each GammaAction ....

....of the GammaAction CA action would be initialised. We have chosen to treat individual roles within an action as units of distribution because this allows them to be executed in the locations at which information is produced or consumed. Although several other approaches view CA actions (e.g. [Romanovsky et al. 1997] [Wellings Burns 1997] as packages (modules, objects, etc. such packages cannot be split into parts and distributed. The general idea of attaching exception handling to roles suits role distribution very well, in spite of the fact that these handlers are controlled class Consumer extends ....

A. Romanovsky, B. Randell, R. Stroud, J. Xu, and A. Zorzo, "Implementation of Blocking Coordinated Atomic Actions Based on Forward Error Recovery ". In Journal of Systems Architecture, 43, pp.687-699, 1997.

....be needed, would have to provide more sophisticated features: aborting any number of nested actions, executing the abortion handlers, signalling exceptions from the nested actions, and involving all participants of the nested actions in the abortion activity, etc. Pre emptive CA action schemes [Romanovsky et al. 1997] provide these features, and can be implemented by the decentralised protocol described in [Romanovsky et al. 1996] The CA action framework allows us to express various time related outcomes of a CA action. If error recovery after a real time exception has been successful within the action, then ....

A. Romanovsky, B. Randell, R. Stroud, J. Xu and A. Zorzo, "Implementation of Blocking Coordinated Atomic Actions Based on Forward Error Recovery," in Journal of Systems Architecture, vol. 43, pp.687-699, 1997.

....the creation of the instances of these actions. This design is very general, so we assume that there is a set of hosts on which the instances of CA actions can be instantiated, and that the scheduler knows their location. We use a centralised CA action scheme, so each action has an action manager [13]. As we have mentioned before, CA actions provide a framework for dealing with exceptions that happen during the execution of the system. In our DSGamma system, when a fault happens inside a GammaAction, a predefined exception ReactionException is raised in the thread executing a role in the ....

ROMANOVSKY, A., RANDELL, B., STROUD, R. J., XU, J., and ZORZO, A. F.: 'Implementation of blocking coordinated atomic actions based on forward error recovery', Journal of System Architecture - Special Issue on Dependable Parallel Computing Systems, 1997, 43(10), pp. 687-699

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