RLH: Bitmap compression technique based on run-length and Huffman encoding

In this paper we propose a technique of compressing bitmap indexes for application in data warehouses. This technique, called run-length Huffman (RLH), is based on run-length encoding and on Huffman encoding. Additionally, we present a variant of RLH, called RLH-N. In RLH-N a bitmap is divided into N-bit words that are compressed by RLH. RLH and RLH-N were implemented and experimentally compared to the well-known word aligned hybrid (WAH) bitmap compression technique that has been reported to provide the shortest query execution time. The experiments discussed in this paper show that: (1) RLH-compressed bitmaps are smaller than corresponding WAH-compressed bitmaps, regardless of the cardinality of an indexed attribute, (2) RLH-N-compressed bitmaps are smaller than corresponding WAH-compressed bitmaps for certain range of cardinalities of an indexed attribute, (3) RLH and RLH-N-compressed bitmaps offer shorter query response times than WAH-compressed bitmaps, for certain range of cardinalities of an indexed attribute, and (4) RLH-N assures shorter update time of compressed bitmaps than RLH.