1st International ICST Conference on Performance Evaluation Methodologies and Tools

Research Article

A general performance model interchange format

  • @INPROCEEDINGS{10.1145/1190095.1190102,
        author={Peter G.  Harrison and Catalina  M. Llado and Ramon  Puigjaner},
        title={A general performance model interchange format},
        proceedings={1st International ICST Conference on Performance Evaluation Methodologies and Tools},
        publisher={ACM},
        proceedings_a={VALUETOOLS},
        year={2012},
        month={4},
        keywords={},
        doi={10.1145/1190095.1190102}
    }
    
  • Peter G. Harrison
    Catalina M. Llado
    Ramon Puigjaner
    Year: 2012
    A general performance model interchange format
    VALUETOOLS
    ACM
    DOI: 10.1145/1190095.1190102
Peter G. Harrison1,*, Catalina M. Llado2,*, Ramon Puigjaner2,*
  • 1: Department of Computing, Imperial College London, 180 Queens Gate, London, SW7 2BZ, UK
  • 2: Dep. de Ciencies, Matematiques i Informatica, Universitat de les Illes Balears, 07071, Palma de Mallorca, Spain.
*Contact email: pgh@doc.ic.ac.uk, cllado@uib.es, putxi@uib.es

Abstract

A common, XML-based interface between specifications of quantitative system models and programmed solutions is developed and illustrated with several examples. It is based on the PMIF (Performance Model Interchange Format), which allows queueing network models to be specified in XML and solved by calling any appropriate modelling tool, such as Qnap. The definition of PMIF specifications is generalised by considering more abstract collections of interacting nodes, using concepts compatible with the Reversed Compound Agent Theorem (RCAT). The interactions are more general in that they synchronise transitions in a pair of nodes rather than being restricted to describing traffic flows. The generalised nodes are characterised by the interactions in which they participate, together with their rates and reversed rates, which may be implicit. In this way, generalised queueing networks with negative customers and triggers can be incorporated and fixpoint models can also be handled uniformly through the use of symbolic variables.