QoSMOS: QoS metrics management tool suite

Researchers are unanimous that the database management system (DBMS) is one of the most promising candidates to be the backbone of quality-conscious information services. Usually, this quality is measured by subjective metrics that can be interpreted as the degree to which a DBMS possesses a given functionality that affects its quality. Note that the Quality of Service (QoS) concerns all phases of the designing database application life-cycle, with a particular interest to the physical design phase - considered as the funnel of other phases. The satisfaction of the QoS in the context of the physical design is ensured by metrics. A metric can be seen as a function returning a single numerical vale whose input parameters belong to (i) the database application (including its schema and queries), (ii) the DBMS hosting that database and (iii) the deployment platform. These metrics are usually expressed by analytical cost models whose development is time consuming and requires calibration to reflect the evolution of the database technology (that refers to both software and hardware). Face to this situation, the presence of management tools dedicated to the construction, exploitation and calibration of cost models becomes a necessity for researchers and students. In this paper, we first propose a domain specific language, called CostDL, to develop cost models for QoS-enabled database physical design. The development of such a language necessitates the explicitation of the different entries of a cost model thanks to meta-modeling techniques. Secondly, to increase their reuse, a persistent repository is proposed to store the new developed cost models. The two contributions are regrouped in a framework called QoSMOS providing functionalities related to the management of cost models.