- Adaptively partition based on query-workload
- Precompute Bloom join between the most frequent triples joins (MF-TJ) combinations
- Partition related properties based on a greedy algorithm and a cost model
- Current version is implemented to run over Apache Spark