next up previous contents index
Next: Software Architect Up: Parameter Model Previous: Structure   Contents   Index

Related Work

Many performance prediction approaches or performance related meta models neglect the influence of parameters values to the above described properties. The UML SPT profile [36] as well as the CSM [37] do not include notions of parameters. Methods that build on these modelling approaches such as CB-SPE [40] thus also cannot express the influence of parameters to QoS properties.

KLAPER [38] allows characterising parameters values, but does not include a formal way of creating abstractions for parameters, because the kind of parameter specification is left open for developers. This limits the use of tools evaluating KLAPER instances, because they can not foresee all possible ways of abstracting parameters. Thus, manual work is required with KLAPER to complete the performance prediction process if parameters are involved.

The ROBOCOP component model [41] also allows characterising parameter values. However, as ROBOCOP aims at embedded systems, it is assumed that the domain for parameter values is very limited. It is possible to model parameters with constant values only, stochastical characterisation for parameter abstractions are not in the scope of that work.

The performance prediction approach by Hamlet et. al. [27] models components as functions and divides their input space into several subdomains. For each subdomain, which can be conceived as a parameter abstraction, different execution times can be determined. However, subdomains are always built for the values of parameters in this approach, other attributes of parameters are neglected.


next up previous contents index
Next: Software Architect Up: Parameter Model Previous: Structure   Contents   Index
Snowball 2007-03-16