The DML Bench is a set of tools that have been developed to work with the Descartes Modeling Language (DML). DML Bench consists of the following parts:
- DML Editors: Textual and graphical editors to simplify the creation and editing of DML models.
- DML Solvers: Solving tools to conduct performance analyses based on DML models.
- S/T/A Adaptation Framework: Domain specific modeling language to describe adaptation processes.
DML provides performance model abstractions for component-based software systems specifically designed for online use. DML Bench provides modeling and prediction facilities that enable online performance prediction during system operation. Performance questions can be answered on the model level, i.e., analyses about the impact of workload changes or system reconfigurations can be conducted without executing expensive performance tests.
For the analysis of DML models, i.e., to conduct performance predictions based on DML models, we provide DML Solvers. The DML Solvers are triggered using performance queries formulated with the Descartes Query Language (DQL). Based on a given performance query, the implemented performance prediction process selects a suitable model solving technique and abstraction level, and returns the requested performance metrics. The currently available DML Solvers use Bounds Analysis and a transformation to Queueing Petri Nets (QPNs) that are then solved with SimQPN which is part of QPME.
S/T/A Adaptation Language and Framework
The adaptation language S/T/A is a domain specific language to describe adaptation processes in dynamic environments (e.g., modern virtualized data centers) based on architecture-level performance models. It is also specified in EMF Ecore.