Jedi: A Storage Manager for SIMD-aware, Worst-case Optimal Join Processing

This talk describes a new graph-pattern engine called Jedi. Using a recent simplification of worst-case optimal join algorithms due to Ngo et al., Jedi translates join queries into a series of set intersection and union operations. Such set operations are ideally suited to modern CPUs that provides single-instruction, multiple data (SIMD) instructions. Using these ideas, we demonstrate that Jedi outperforms specialized graph engines by over an order of magnitude and relational systems by over two orders of magnitude on standard graph processing queries over real data.