Combine the churn fluctuation model with depth reduction

Establish a rigorous combination of the deletion-churn fluctuation model—including its Gamma-distributed offered-load approximation and failure-rate estimate—with the multitable depth-reduction techniques of Section 3, in order to determine whether the resulting churn-safety predictions remain valid for cascades whose tails have been shortened.

Background

The paper analyzes continuous deletion-and-insertion churn using a mean-field loss-system model. It supplements that model with heuristic fluctuation corrections, including a variance recursion, level-one feedback, and a Gamma closure for the offered load at deeper levels. These approximations are used to estimate the terminal-overflow rate and to identify churn-safe load-factor targets.

Separately, the multitable construction uses progressive overfill and tail truncation to reduce cascade depth and negative-lookup cost. The paper does not establish how these depth-reduction mechanisms interact with the heuristic churn model, especially when the shortened cascade has a nearly empty final level whose failure behavior is sensitive to fluctuations. Consequently, a rigorous or validated analysis of this combined configuration remains unresolved.

References

The model is pessimistic by the factor just quoted, it does not model the fill transient, and the Gamma tail is an assumption that carries $D$ when the last level is nearly empty; its combination with the depth reduction of Section~\ref{sec:depth} is open.

— MultiTable: A Faster Hash Table at any Physical Load Factor up to and Including One  (2609.39233 - Petkus, 30 Sep 2026) in Appendix, Section “Deletion churn: model and measurements,” paragraph “Fluctuations and validation”