There are many challenges of data sharing both in grid environment and traditional data integration domains, such as sharing between heterogeneous, distributed data sources. In grid environment, there are many specific features: limited data sharing, frequently changing data sources and large amounts of concurrent users. Under these conditions, it is strongly needed to provide more usable, more flexible system software and unique application programming interfaces to solve these problems. Vega Information Grid team has provided REVP data model and invented a flexible schema mapping tool employing reference technique to achieve these goals. In addition, performance in grid environment is especially important and needs some different optimization strategy. VIG has taken some optimization measures to improve the performance for distributed query. This paper elaborates the REVP data model and virtual database engine. Performance test shows that the query optimization measures are effective.