Application-independent reconciliation for nomadic applications - Archive ouverte HAL Access content directly
Conference Papers Year : 2000

Application-independent reconciliation for nomadic applications

(1) , (2) , (2)
1
2
Marc Shapiro
Anne-Marie Kermarrec
  • Function : Author
  • PersonId : 830783

Abstract

We describe attempts to build an application-independent model to support reconciliation of diverged replicas of shared objects. While replicas are disconnected from one another, actions on the shared objects are recorded in a log. An action is composed of a precondition, an operation and a postcondition. When reconnecting, the system attempts to reconcile the divergent replicas, in several phases. A symbolic phase merges the separate logs, creating one or more schedules, such that preconditions from one log remain true despite the postconditions introduced by the other. Then, a simulation phase checks the possible outcomes by actually applying the schedules to scratch copies of the shared objects. Finally, a selection phase allows users or applications to select one of the candidate schedules.Our approach supports sharing general objects, where one update may reference multiple objects (not just a single file or database). Compared to previous work on log-based reconciliation, our logs capture more semantic information and provide the application with more powerful and finer control over the outcome of reconciliation
Fichier principal
Vignette du fichier
sigops-ew-2000-logmerge.pdf (179.96 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

inria-00444646 , version 1 (07-01-2010)

Identifiers

  • HAL Id : inria-00444646 , version 1

Cite

Marc Shapiro, Antony Rowstron, Anne-Marie Kermarrec. Application-independent reconciliation for nomadic applications. 9th ACMSIGOPS European Workshop 2000 sigopsew :``Beyond the PC:New Challenges for the Operating System'', sigops, 2000, Kolding, Denmark, Denmark. ⟨inria-00444646⟩
122 View
141 Download

Share

Gmail Facebook Twitter LinkedIn More