The metadata in the Semantic Grid are described using the technologies of the Semantic Web, such as OWL (OWL Web Ontology Language), and SPARQL has been recommended by W3C. And most metadata including actual data are stored in relational databases. Under such a situation, the necessity of SPARQL-to-SQL translation algorithm development becomes issued for seamless data access to the Semantic Grid data. Even though several translation algorithms have been proposed, there still remain the following problems, and especially these algorithms are dependent on a specific storage model. Hence, when we try to use one of other storages not the original, much cost is required to modify and verify the translation algorithm. In this paper, we propose a new system model to independently use an algorithm upon storage structures. The system model guarantees independency between translation algorithms and storages, and thus it improves application and usability of translation algorithms.