Lockless hash tables with low false negatives

Hash tables are efficient storage data structures widely used in many types of high-performance computer-related problems. In their design, optimal trade-offs must be made to accommodate for the specific characteristics of the application. In this paper we present lock-free low-false-negative (LFN) tables, a family of hash tables designed to address one such type of trade-off. LFN tables sacrifice a low probability of false negatives and a very low (or negligible) probability of false positives to achieve higher performance access time in concurrent shared memory applications. LFM tables are structurally biased towards false negatives and therefore are more suitable for applications that tolerate better false negatives than positives. In this paper we provide a mathematical analysis of their performance and provide use cases where they can be deployed to eliminate shared memory access bottlenecks in the context of very high-speed computer networks.