Semi-Naive Query Planning For Grid Data Integration

We consider the problem of querying multiple grid data sources under the assumption that quantitative metadata of data access costs is not available. In this case, query planning must rely on qualitative information to improve performance. We introduce a semantic approach that allows the hand-crafting of qualitative data source characterizations related to access costs. Our technique may offer substantial optimization opportunities in practical situations. In addition, it can uniformly describe data sources with reduced query capabilities, such as data-producing computations and devices.