Concurrency Control and Parallelism Advances

The field of concurrency control and parallelism is moving towards innovative solutions that enable efficient and scalable processing of high-contention workloads. Recent developments focus on designing novel protocols and data structures that can handle insert-dominant workloads, predict transaction conflicts, and provide deadlock-free transaction execution. Notable advancements include the use of predictive scheduling, multi-version-based optimization, and hybrid morsel dispatching policies. These approaches have shown significant improvements in throughput, commit rates, and concurrency, demonstrating the potential for substantial performance gains in various applications. Noteworthy papers include: PIPQ, which introduces a strict and linearizable concurrent priority queue with parallel insertions, and ForeSight, which presents a high-performance deterministic database system with predictive scheduling and conflict-aware schedules. Brook-2PL is also notable for its novel two-phase locking protocol that tolerates high contention workloads, and the work on robust recursive query parallelism in graph database management systems presents a hybrid policy that captures the behavior of both source and frontier morsel-only policies.

Sources

PIPQ: Strict Insert-Optimized Concurrent Priority Queue

ForeSight: A Predictive-Scheduling Deterministic Database

Brook-2PL: Tolerating High Contention Workloads with A Deadlock-Free Two-Phase Locking Protocol

Robust Recursive Query Parallelism in Graph Database Management Systems

Built with on top of