8th IEEE International Conference on Collaborative Computing: Networking, Applications and Worksharing

Research Article

CUERA: Performance evaluation of a generic data- and undo/redo-consistency framework

Download94 downloads
  • @INPROCEEDINGS{10.4108/icst.collaboratecom.2012.250374,
        author={Daniel Stolzenberg and Erika M\'{y}ller},
        title={CUERA: Performance evaluation of a generic data- and undo/redo-consistency framework},
        proceedings={8th IEEE International Conference on Collaborative Computing: Networking, Applications and Worksharing},
        publisher={IEEE},
        proceedings_a={COLLABORATECOM},
        year={2012},
        month={12},
        keywords={computer supported collaborative work (cscw) optimistic replication eventual consistency commutative replicated data type (crdt) any-undo/redo performance evaluation},
        doi={10.4108/icst.collaboratecom.2012.250374}
    }
    
  • Daniel Stolzenberg
    Erika Müller
    Year: 2012
    CUERA: Performance evaluation of a generic data- and undo/redo-consistency framework
    COLLABORATECOM
    ICST
    DOI: 10.4108/icst.collaboratecom.2012.250374
Daniel Stolzenberg1,*, Erika Müller1
  • 1: Institute of Communications Engineering - University of Rostock
*Contact email: daniel.stolzenberg@uni-rostock.de

Abstract

In this paper, a generic data- and undo/redo-consistency framework for realtime interactive collaboration applications is evaluated in terms of performance and scalability. Since data is represented based on Entity-Relationship-Models, the framework is applicable in a wide range of domains. It provides a set of state-focusing operations, that is able to abstract arbitrary data interfaces. In addition, non-reflexive meta-operations allow to undo/redo any user action. Consistency is maintained by forcing concurrent operations to commute by transitive precedence rules. Both convergence and undo/redo control are mirrored at runtime to provide maximal performance and scalability regarding session duration, data and number of participants. These claims are successfully verified in this paper by analyzing average execution times under a realtime collaboration workload simulated in a turn-based system.