Reimplementation of Lock Free Cuckoo Hashing

This paper is presenting a lock-free cuckoo hashing algorithm. The algorithm allows modifying operations to operate as one with search ones and requires only single word compare-and-swap primitives. Searching of items can operate concurrently with others mutating operations, thanks to this two-round query protocol that has been enhanced with a logical clock technique. When an insertion triggers a sequence of key displacements, rather than locking the path, the algorithm will break down the relocation sequence into single relocations which can be executed independently and concurrently with other operations. A finely tuned synchronization and a helping mechanism for relocation are designed. The mechanisms will allow high concurrency and provide progress guarantees for the data structure's operations. This should represent an upgrade over current hashing algorithms which do not allow for concurrent searching and mutating operations with hash tables. The result shows that this lock-free cuckoo hashing system performs consistently better than two efficient lock-based hashing algorithms, the chained and the hopscotch scenarios.

Cuckoo Feature Hashing: Dynamic Weight Sharing for Sparse Analytics

Feature hashing is widely used to process large scale sparse features for learning of predictive models. Collisions inherently happen in the hashing process and hurt the model performance. In this paper, we develop a feature hashing scheme called Cuckoo Feature Hashing(CCFH) based on the principle behind Cuckoo hashing, a hashing scheme designed to resolve collisions. By providing multiple possible hash locations for each feature, CCFH prevents the collisions between predictive features by dynamically hashing them into alternative locations during model training. Experimental results on prediction tasks with hundred-millions of features demonstrate that CCFH can achieve the same level of performance by using only 15%-25% parameters compared with conventional feature hashing.