PaMeCo join: A parallel main memory compact hash join
作者：
Highlights：
• We developed a memory-constrained hash join suitable for column-store main-memory databases.
• We utilize a block nested loops approach, performing a hash join between blocks that uses a compact hash table to lower the memory footprint.
• We tackle the challenges of parallelizing the join algorithm while being mindful of operating within a tight memory constraint.
• In scenarios without memory constraints our solution performs competitively with other contemporary non-hardware tuned join algorithms.
• In memory-constrained scenarios, our solution is up to 3 times faster than another high-performance memory-constrained hash join.
摘要
Highlights•We developed a memory-constrained hash join suitable for column-store main-memory databases.•We utilize a block nested loops approach, performing a hash join between blocks that uses a compact hash table to lower the memory footprint.•We tackle the challenges of parallelizing the join algorithm while being mindful of operating within a tight memory constraint.•In scenarios without memory constraints our solution performs competitively with other contemporary non-hardware tuned join algorithms.•In memory-constrained scenarios, our solution is up to 3 times faster than another high-performance memory-constrained hash join.
论文关键词：In-memory databases,Hash join,Memory constrained,OLAP,Column store
论文评审过程：Received 31 August 2015, Accepted 5 October 2015, Available online 6 November 2015, Version of Record 30 April 2016.
论文官网地址：https://doi.org/10.1016/j.is.2015.10.004
