Optimization of Compact Set Membership Representation for Distributed Computing March 2005

Bloom filters are widely used although not always optimal. Notably in application areas where false negatives are bearable, other techniques can clearly be better. This master thesis shows that at least for a specific area in the parameter space Bloom filters are significantly outperformed even by trivial methods. We provide an analysis, show that many application areas where Bloom filters are deployed do not require the strong ”no false negatives” policy, and also examine application specific issues in a distributed web caching scenario. We believe we have opened up a new perspective with which to approach the subject and show that the unconditional use of Bloom filters is questionable.