Core Concepts
Aleph Filter is an infinitely expandable filter that supports all operations (insertions, queries, deletes) in constant time, while providing far superior memory vs. false positive rate trade-offs compared to existing methods.
Abstract
The content discusses the design and analysis of Aleph Filter, an infinitely expandable filter that addresses the limitations of existing expandable filters.
Key highlights:
Aleph Filter supports all operations (insertions, queries, deletes) in constant time, regardless of how much the data grows. This is achieved by duplicating void entries (entries that have run out of fingerprint bits) across the main hash table.
Aleph Filter provides a superior memory vs. false positive rate (FPR) trade-off compared to existing methods. This is done by pre-allocating slightly longer fingerprints from the onset and assigning decreasing fingerprint lengths to newly inserted entries. Even if the data size surpasses the initial estimate, Aleph Filter continues to outperform the state-of-the-art.
Aleph Filter addresses the challenge of efficiently supporting deletes by transforming void entries into tombstones and lazily identifying and removing their duplicates before the next expansion.
The analysis shows that the fraction of duplicated void entries stays moderate, and therefore does not significantly impact the FPR or the maximum number of expansions that Aleph Filter supports.