C. Hofmeister and J. Purtilo. Dynamic Reconfiguration in Distributed Systems-Adapting software modules for replacement. In Proceedings of the 13th International Conference on Distributed Computing Systems, pages 101-- 110, Pittsburgh, May 1993. IEEE Computer Society Press.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....are located fail. 5. The machine on which the RSS daemon is executing must be failure free for the duration of the application. 7 Related Work Checkpointing parallel applications have been widely studied in [16, 29, 25] and checkpointing systems for parallel applications have been developed [12, 10, 33, 38, 31, 15, 20, 34, 3, 23, 20, 4, 22, 21, 27]. Some of the systems were developed for homogeneous systems [12, 11,33,34] while some checkpointing systems allows applications to be checkpointed and restarred on heterogeneous systems [15, 20, 3 5, 23, 21, 27] Calypso [5] and Plinda [23] require application writers to write their programs in ....

....and can only be used with PVM environments. Cocheck [34] and Starfish [3] provide fault tolerance with their own MPI implementations and hence are not suitable for distributed computing and Grid systems where the more secure MPICH G [19] is used. CUMULVS [20] Dome [4, 7] the work by Hofmeister [22] and Deconick [13, 14, 9] DRMS [27] and DyRecT [2, 21] are closely related to our research in terms of the checkpointing API, the migrating infrastructure and reconfiguration capabilities. The CUMULVS [20] API is very similar to our API in that it require the application writers to specify the ....

[Article contains additional citation context not shown here]

C. Hofmeister and J. M. Purtilo. Dynamic Reconfiguration in Distributed Systems : Adapting Software Modules for Replacement. In Proceedings of the 15 th International Conference on Distributed Computing Systems, Pittsburgh, USA, May 1993.

....known as dynamic module replacement that enables the implementation of a component to be hot swapped at runtime. Existing solutions have various shortcomings that limit their scope of applicability and deployment. For example, the approaches taken by CONIC [12] ARGUS [2] and POLYLITH [8] are coupled to particular research languages that generally lack support for mainstream software development. Other solutions have targeted development languages, but at a cost: HADAS [1] for Java) requires integrated tool support at runtime; dynamic Java classes [13] compromise portability by ....

....3 Module Replacement with Serfs This section outlines five sufficient factors for supporting dynamic module replacement, and describes how Serfs satisfy these prerequisites. Previous research has explored other factors pertaining to the safety and semantic correctness of module replacement [9, 8]. A mature reconfiguration strategy should address these additional factors such as protection, security, and substitution consistency. Ultimately, however, all approaches must supply a sufficient infrastructure for effecting the module replacement itself. Our concern is to supply such a ....

C. Hofmeister and J. M. Purtilo. Dynamic reconfiguration in distributed systems: Adapting software modules for replacement. In Intl. Conf. on Distributed Computing Systems, pages 101--110, 1993.

.... fields of use [Milojicic99] Ambler99] Application mobility can be used to dynamically balance the load between several machines in a distributed system [Nichols87] to reduce network traffic by moving clients closer to servers [Douglis92] to dynamically reconfigure distributed applications [Hofmeister93], to implement mobile agent platforms [Chess95] or as a machine administration tool [Oueichek96] Application persistence can be used for fault tolerance [Wojcik95] or for application debugging. Distributed applications development is an important research direction in computing systems. In this ....

C. Hofmeister and J. M. Purtilo. Dynamic Reconfiguration in Distributed Systems : Adapting Software Modules for replacement. Proceedings of the 13 th International Conference on Distributed Computing Systems, Pittsburgh, USA, May 1993.

.... the approaches are often suited to a specific application or domain; they are based on particular frameworks (e.g. CORBA) or programming language (e.g. Java [MG99] they only deal with certain classes of systems (e.g. client server [Kin93] or reconfiguration (e.g. replacement by a copy [HP93] Many of the current approaches follow the Configuration Programming philosophy of Kramer and colleagues. Its principles are briefly stated in [KM98] 1. The configuration language used for structural description should be separate from the programming language used for basic component ....

Christine Hofmeister and James Purtilo. Dynamic reconfiguration in distributed systems: Adapting software modules for replacement. In Proceedings of the 13th International Conference on Distributed Computing Systems, pages 101--110, Pittsburgh, May 1993. IEEE Computer Society Press.

....of RESTCLK. These examples illustrate the use of observation and control for transparent dynamic reconfiguration of application systems in various situations. We also use these examples to compare RESTCLK solutions for dynamic reconfiguration with solutions proposed in other systems like polylith [19, 38, 21, 22, 20, 5] and conic [25, 27, 26, 24] These two are the only other significant systems that we are aware of, which propose novel communication systems organization that facilitate dynamic system reconfigurations of the kinds discussed here. The examples discussed in this chapter will illustrate clearly the ....

....received and sent by an object. RESTCLK compiler may be written to translate these statements automatically to the requisite control signal emissions. 9.4 Other Related Work Work on dynamic reconfiguration may be classified into three categories: 1. Problem identification and proposed solutions [24, 25, 15, 20, 21, 19, 44, 17, 18, 41, 34], 2 We show in Chapter 8 programming an object to function in a RESTCLK environment will in practice cost very little in terms of additional delays incurred in communication, when compared to any other agent mediated communication system. Thus there is little or no overhead to using RESTCLK. ....

C. Hofmeister and J. Purtilo. Dynamic reconfiguration in distributed systems: Adapting software modules for replacement. In Proceedings of the International Conference on Distributed Computer Systems, pages 101--110, May 1993.

.... this model because, as stated in [2] other work in dynamic reconfiguration either only deals with tool and language support to describe and execute the changes [1] assuming that the system is already in a safe state, or it imposes limitations to guarantee maintenance of the system s state [4, 3, 9]. Limitations can be on the kind of system (e.g. only client server) or on the kind of changes handled (e.g. a component may only be replaced by a specialization of it) Sometimes they are due to the existence of special mechanisms to capture and recover the application state prior to the ....

C. Hofmeister and J. Purtilo. Dynamic reconfiguration in distributed systems: Adapting software modules for replacement. In Proceedings of the 13th International Conferenceon Distributed Computing Systems, pages 101--110, Pittsburgh, May 1993. IEEE Computer Society Press.

....independent package has also been developed by Silva, Veer, and Silva [24] who have created a library based system where the user is responsible for inserting calls to specify the data to be saved and perform the checkpoints. Another system related to ours was developed by Hofmeister and Purtilo [18]. As in Dome, they use a preprocessing mechanism for saving the state of distributed programs. 7. Future work The Dome system is undergoing very active development. We are adding new classes and working with both computer science and computational science researchers to develop more production ....

C. Hofmeister and J. Purtilo. Dynamic reconfiguration in distributed systems: Adapting software modules for replacement. Technical Report UMIACS-TR-92-120, University of Maryland, November 1992.

....condition always holds when the middleware adaptation is due to environmental changes. On the other hand, adaptation due to new non functional requirements must be dealt with on a case by case basis. The latter condition requires the middleware to be able to import and export its state (e.g. [Hofmeister93]) a function that can be directly supported by reflective middleware (e.g. see [Killijian99] Note that meeting this condition is specific for each pair of middleware configurations involved in the exchange (regarding implementation of the state mapping function) We cannot afford the definition ....

Hofmeister, C., J. Purtilo, "Dynamic Reconfiguration for Distributed Systems, Adapting Software Modules for Replacement", Proceedings of the 13th IEEE International Conference on Distributed Computing Systems (ICDCS'93), May 1993.

....modules which can be reconfigured in this manner were identified. ffl The module s state can be safely discarded during reconfiguration. ffl The module does not require any special initialization when created dynamically. ffl There is no synchronization between the module and its neighbors. In [HP93] a tool was described to capture and restore a module s state automatically. The approach used was to preprocess the module code to insert statements to capture and restore the state. However, since the saved state contains values of variables of the program and also explicit reference to ....

C. Hofmeister and J. Purtillo. "Dynamic reconfiguration in distributed systems: adapting software modules for replacement". In Proc. 13th International Conference on Distributed Computing Systems, pages 101-- 111, Pittsburg, May 1993.

....## ## ## A2 moves fromnode 2 to node 3, message m3 is still in transit Figure 2 Status of communication channels 2 Related work Different approaches can be considered for dealing with potential inconsistencies when reconfiguration occurs. Reaching a reconfiguration point The Polylith [7][10] environment provides support for the reconfiguration of distributed applications. Applications are designed as a set of modules interconnected through a message bus. Polylith allows module reconfiguration from different points of view : structural, geometric or implementation changes. The ....

C. Hofmeister, J. Purtilo, "Dynamic Reconfiguration in Distributed Systems : Adapting Software Modules for replacement", Proc. 13th International Conference on Distributed Computing Systems, pp. 101-110, 1993.

....about the time to reach it. Rebinding of agents can be dynamically done by virtue of the concept of roles. The problems of storing incoming events in order to be able to replay them when the agents are rebound is greatly facilitated by the use of a single communication bus, like in Polylith [Hofmeister93] In addition, the causal ordering of events in our communication bus helps in maintaining the communication consistency of the reconfigured agents and the rest of the application. The validation of reconfiguration actions is provided on one hand by the architecture description which specifies ....

C. Hofmeister, J. Purtilo, "Dynamic Reconfiguration in Distributed Systems : Adapting Software Modules for replacement", Proc. 13th International Conference on Distributed Computing Systems, pp. 101-110, 1993.

....package has also been developed by Silva, Veer, and Silva [29] who have created a purely library based system where the user is responsible for inserting calls to specify the data to be saved and perform the checkpoints. Another system related to ours was developed by Hofmeister and Purtilo [19]. As in Dome, they use a preprocessing mechanism for saving the state of distributed programs. While their main concern is dynamic program reconfiguration rather than checkpoint and restart, their preprocessing method is similar to the one we are using. Finally, Duda [8] has analyzed the expected ....

Christine Hofmeister and James Purtilo. Dynamic reconfiguration in distributed systems: Adapting software modules for replacement. Technical Report UMIACS-TR-92-120, University of Maryland, November 1992.

....as persistent state during reconfiguration. This relies heavily on the use of strict subtyping as otherwise replacements nodes will not be able to process queued messages. Hofmeister and Purtillo address the problem of automatically capturing and reinstating application specific state at run time [9]. The method presented is suited to exact copy replacement of nodes and heterogeneous processes migration, but does not deal with other types of reconfigurations. 3 A Safe State for Reconfiguration We assume that distributed applications are composed of a finite number of nodes which can ....

C. Hofmeister and J. Portilo. Dynamic reconfiguration in distributed systems : Adapting software modules for replacement. In 13th International Conference on Distributed Computing Systems, pages 101--110. IEEE, May 1993.

....processes stacks and heaps are not saved) In contrast, in [7] the programming paradigm used for writing applications (coarse grained dataflow) is used to keep the memory overhead for checkpoints low. Studies on application reconfiguration and heterogeneous process migrations are described in [11] and [20] A method for automatically saving and restoring the application stack is presented in [11] Similarly to our work, the programmer explicitly specifies the points were the application can be reconfigured (e.g. moved to another machine) The source program is then transformed so as to ....

....writing applications (coarse grained dataflow) is used to keep the memory overhead for checkpoints low. Studies on application reconfiguration and heterogeneous process migrations are described in [11] and [20] A method for automatically saving and restoring the application stack is presented in [11]. Similarly to our work, the programmer explicitly specifies the points were the application can be reconfigured (e.g. moved to another machine) The source program is then transformed so as to keep track of the structure of the application stack. In contrast to the work described in [11] ....

[Article contains additional citation context not shown here]

Hofmeister, C., and Purtilo, J. Dynamic reconfiguration in distributed systems: Adapting software modules for replacement. In Proc. of 13th International Conference on Distributed Computing Systems (Pittsburgh, Pennsylvania, May 1993), pp. 101--110.

....be a cheap operation, and we want to take advantage of this fact to concentrate on other issues. We are mainly concerned with using the object oriented paradigm to maximize the user transparency of a fully portable checkpoint and restart mechanism for distributed programs. Hofmeister and Purtilo [7] have written a preprocessor for saving the state of programs written using their Polylith system. While their main concern is dynamic program reconfiguration rather than checkpoint and restart, their preprocessing method is somewhat similar to the one we describe below. 3 The Model It is ....

Christine Hofmeister and James Purtilo. Dynamic reconfiguration in distributed systems: Adapting software modules for replacement. Technical Report UMIACS-TR-92-120, University of Maryland, November 1992.

....the module s output should be lost or duplicated as a result of the migration. These problems imply certain requirements for distributed environments that support reconfiguration. Hofmeister and Purtilo put forth the following requirements for reconfiguration in heterogeneous distributed systems [8] 1 : R1) communication across heterogeneous hosts (R2) current configuration is accessible (R3) bindings (interconnections) are not compiled into modules (R4) no covert communication among modules (R5) ability to add remove modules and bindings (R6) access to messages in transit (R7) mechanism ....

....that are used to automatically prepare a process 1 This set of requirements does not appear in their paper, but was part of the conference presentation. to participate during reconfiguration and special techniques are used to capture internal program state in order to accomplish the migration [8]. In this paper, we consider logical and physical dynamic reconfiguration in heterogeneous distributed systems whose modules are written using a new programming model called I O abstraction [5] Briefly, I O abstraction is the view that each software module in a system has a set of data ....

Christine R. Hofmeister and James M Purtilo. Dynamic reconfiguration in distributed systems: Adapting software modules for replacement. In Proceedings of the 13th International Conference on Distributed Computing Systems, Pittsburgh, Pennsylvania, pages 101--110, May 1993.

....Their approach suffers when the application is communication bound because it doubles the number of data messages. In contrast Dome s mechanism takes a more conventional approach by saving the checkpoints to stable storage rather than relying on shadow processes. Hofmeister and Purtilo [12] have written a preprocessor for saving the state of programs which use their Polylith system. While their primary focus is dynamic program reconfiguration rather than checkpoint and restart, their preprocessing method is somewhat similar to the one described here. 7 Future Work There are a ....

Christine Hofmeister and James Purtilo. Dynamic reconfiguration in distributed systems: Adapting software modules for replacement. Technical Report UMIACS-TR-92-120, University of Maryland, November 1992.

No context found.

C. Hofmeister, J. Purtilo, "Dynamic Reconfiguration in Distributed Systems: Adapting Software Modules for Replacement," Proceedings of the IEEE 13th International Conference on Distributed Computing Systems, pp. 101-110, May 1993.

No context found.

C. Hofmeister and J. Purtilo. Dynamic Reconfiguration in Distributed Systems-Adapting software modules for replacement. In Proceedings of the 13th International Conference on Distributed Computing Systems, pages 101-- 110, Pittsburgh, May 1993. IEEE Computer Society Press.

No context found.

C.R. Hofmeister, and J.M. Purtilo, "Dynamic Reconfiguration in Distributed Systems: Adapting Software Modules for Replacement", in 13th International Conference on Distributed Computing Systems, May 1993.

No context found.

Hofmeister C, Purtilo JM. Dynamic reconfiguration in distributed systems: Adapting software modules for replacement. 13th International Conference on Distributed Computing Systems (ICDCS'93), Pittsburgh, PA, May 1993.

No context found.

Springer. Hofmeister, C and Purtilo, J (1993). Dynamic reconfiguration in distributed systems: Adapting software modules for replacement. In Proc, Conf on Distributed Computing Systems. IEEE.

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