
215Scalable End-User Access to Big Data
To address efciency, we propose to decompose the “Query Answering”
component into several layers, as shown in Figure 6.2:
1. Query Transformation using the ontology and mappings
2. Query Planning to distribute queries to individual servers
3. Query Execution using existing scalable data stores, or a massively
parallelized (cloud) architecture
The implementation of the query transformation layer can take recent
theoretical advances in query rewriting into account, which can lead to sig-
nicantly improved performance (see section “Query Transformation”). The
same holds for query execution, which ...