Upscaledb: Efficient integer-key compression in a key-value store using SIMD instructions
作者：
Highlights：
•
• We apply a range of integer compression techniques to the compression of keys to an open source production-quality B+-tree.
• We assess experimentally the effect of compression on various operations (insertions, cursors, sums, etc.).
• We find that the best techniques (SIMD FOR and BP128) can both reduce memory usage and improve analytical performance.
• The best techniques leverage the SIMD instructions of modern commodity processors.
摘要
Highlights •We apply a range of integer compression techniques to the compression of keys to an open source production-quality B+-tree.•We assess experimentally the effect of compression on various operations (insertions, cursors, sums, etc.).•We find that the best techniques (SIMD FOR and BP128) can both reduce memory usage and improve analytical performance.•The best techniques leverage the SIMD instructions of modern commodity processors.
论文关键词：B+-tree,Data compression,Vectorization,Key-value stores
论文评审过程：Received 28 December 2015, Revised 16 November 2016, Accepted 4 January 2017, Available online 7 January 2017, Version of Record 14 January 2017.
论文官网地址：https://doi.org/10.1016/j.is.2017.01.002
