| Cardinale Y, Hernandez E. Checkpointing facility on a metasystem. European Conference on Parallel Computing (EuroPar ' |
....of C source code. Our prototype is freely available at the following URL: http: sardes.inrialpes.fr research JavaThread It has been successfully used in the Suma metacomputing platform as a basic system for the implementation of a checkpointing recovery facility for parallel computations [9]. In addition to Suma s designers, there were about 200 downloads from users, testers, students and researchers all around the world. #### ######## #### ###### ######### 7 ### ####### The rest of the report is structured as follows. Section 2 rst presents the Java Virtual Machine s ....
....thread migration system. We restricted our discussion to the design and implementation issues in a local environment (i.e. a local JVM) and we did not discuss the problems rising from using our migration facility for building large distributed systems. Some elements of response are presented in [9], where the authors describe the building of a distributed fault tolerant system using our Java thread checkpoint recovery mechanism. Indeed, Cardinale et al. describe their implementation of acheckpoint restart facility for parallel computations in the Suma metacomputing system. The fault ....
Cardinale, Y., Hernndez, E., Checkpointing Facility on a Metasystem, ######## #### ################# ######### ###############, Manchester, UK, January 2001. http://suma.ldc.usb.ve/
No context found.
Y. Cardinale and E. Hernandez. Checkpointing facility in a metasystem. Submitted to Euro-Par
....Los Agentes de Ejecuci on proveen un m etodo unico llamado execute 1 . Un Agente de Ejecuci on recibe una orden de un Representante para ejecutar una aplicaci on, entonces comienza a cargar clases 1 Estamos trabajando actualmente en un servicio de recuperaci on basado en puntos de chequeo [13], lo que incluir a un m etodo llamado resume. 4 y archivos din amicamente. Para las plataformas paralelas, el Agente de Ejecuci on juega el rol de puerta de entrada. El m etodo execute tiene cuatro par ametros: RU execute (in EU, in ExecAgentConf, in AbstractClient, inout ProfilerConf) El ....
....y per laci on, los cuales tienen que ser invocados expl citamente por el usuario. Un Agente de Ejecuci on que provee servicios de recuperaci on con puntos de chequeo debe ejecutar una versi on extendida de la m aquina virtual Java que implementa los puntos de chequeo al nivel de hebras [14] [13]. El proceso de recuperaci on es controlado transparentemente por el Representante que se encarga de la ejecuci on. Un Agente de Ejecuci on que provee servicios de per laci on ejecuta las aplicaciones bajo el control de un per lador, que nosotros llamamos back end pro ler, y retorna una ....
Y. Cardinale and E. Hernandez, \Checkpointing facility in a metasystem," Submitted to Euro-Par
....are File, FileDescriptor, FileInputStream, InputStream (for System.in) and OutputStream (for both System.out and System.err ) All methods de ned in the Client Stub are called by applications running on remote nodes. 1 We are currently working on a checkpointing and recovery facility for suma [8], which will include a method resume. Lecture Notes in Computer Science 5 2.4 Proxy A Proxy has basically two functions. First, it connects a client with an Execution Agent in such a way that these components communicate directly with each other. The second function of a Proxy is to implement ....
....the client s side. Optionally, an Execution Agent may o er checkpointing and pro ling services, which have to be invoked explicitly by the user. An Execution Agent that provides checkpointing and recovery should be able to execute an extended JVM that implements checkpointing at the thread level [5, 8]. The recovery process is controlled transparently by the Proxy in charge of the execution. An Execution Agent that provides the pro ling service executes the application under the control of a pro ler, which we call the back end pro ler, and returns a pro le data structure to the client machine. ....
Y. Cardinale and E. Hernandez. Checkpointing facility in a metasystem. Submitted to Euro-Par
No context found.
Cardinale Y, Hernandez E. Checkpointing facility on a metasystem. European Conference on Parallel Computing (EuroPar '
No context found.
Cardinale, Y., Hernndez, E. Checkpointing Facility on a Metasystem. European Conference on Parallel Computing (Euro-Par'2001.
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