Scholarlink

We demonstrate PJDL, a parallel join system that transforms Leapfrog Triejoin’s depth-first execution into breadth-first processing via incremental result trie construction and a dynamic grouping strategy for multi-way join tasks. PJDL addresses key challenges of LFTJ, including limited parallelism, load imbalance from data skew, and redundant computation during parallel execution. Experiments show that PJDL outperforms PostgreSQL 10.23 and ADOPT across IMDB, TPC-H, and JCCH benchmarks.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

PJDL: Parallelizing Leapfrog Triejoin via Incremental Trie Construction and Dynamic Load Balancing

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jiaqi Wei,
  • Yipeng Liu,
  • Yuming Lin

摘要

We demonstrate PJDL, a parallel join system that transforms Leapfrog Triejoin’s depth-first execution into breadth-first processing via incremental result trie construction and a dynamic grouping strategy for multi-way join tasks. PJDL addresses key challenges of LFTJ, including limited parallelism, load imbalance from data skew, and redundant computation during parallel execution. Experiments show that PJDL outperforms PostgreSQL 10.23 and ADOPT across IMDB, TPC-H, and JCCH benchmarks.