Memory Efficient Parallel Bloom Filters for String Matching

Bloom filter is a fast and simple structure for string matching, but it allows false positives. Longest prefix matching bloom filters (LPM) with counter and extended fast hash table can eliminate false positives at the expense of double memory and none wire-speed. Faster searching speed can be achieved with parallel LPMs (PLPM), using more memory. We present an algorithm named memory efficient parallel bloom filters (MEPBF) which will save (q-1)/(2q) memory and reach nearly the same performance compared with PLPM of q LPMs. With this new algorithm, either faster searching speed can be achieved at given memory and strings or less memory is need at given speed and strings.