Universal superiority of batch-based execution across all queries

Determine whether the batch-based vectorized query execution model in Stardog (BARQ) can ever outperform the legacy tuple-at-a-time Volcano-style execution engine on every query, or whether some query classes inherently resist universal performance gains from batch-based execution.

Background

In discussing why BARQ cannot simply replace the legacy executor in all cases, the paper lists several obstacles to a wholesale switch. The third obstacle explicitly questions whether a batch-based approach will ever dominate across all queries.

The authors note that selective, IO-bound queries are prone to overfetching under batching, which may limit the ability of a vectorized engine to outperform a tuple-at-a-time approach universally. This uncertainty motivates maintaining both execution strategies and selecting between them based on workload characteristics.