Accurate Counting Bloom Filters for Large-Scale Data Processing

Table 1

Reduce-side join performance comparisons in Hadoop.

Filter parameters 𝑚/𝑛 = 10, 𝑘=3	False positive probability	Map inputs (MB)	Map outputs (MB)	Filter construction times (s)	Total execution times (s)
Join + CBF	0.01772	252.5	45.1	62	115
Join + ACBF𝑂	0.00491	252.5	29.8	68	92