Access-Adaptive Priority Search Tree

TL;DR

The paper introduces the access-adaptive priority search tree (AAPST), a data structure that guarantees worst-case O(log n) query time while adapting to data distribution, suitable for real-time applications with strict timing and energy constraints.

Contribution

The AAPST provides distribution-sensitive performance with strict worst-case bounds, improving over splay trees for real-time, energy-efficient applications.

Findings

AAPST achieves O(log n) worst-case query time.

AAPST adapts to data distribution similar to splay trees.

Suitable for real-time systems with strict timing constraints.

Abstract

In this paper we introduce the notion of explicit worst-case bounded adaptive algorithms for applications with fixed process-completion requirements. Such applications demand that a process be guaranteed to complete within an established time interval while adaptively reducing computational overhead during that interval, e.g., so as to reduce total energy usage. Our principal contribution is the access-adaptive priority search tree (AAPST), which can provide efficient distribution-sensitive performance comparable to the splay tree, but do so within strict - and O(logn) optimal - worst-case per-query bounds. More specifically, while the splay tree is conjectured to offer optimal adaptive amortized query complexity, it may require O(n) for individual queries, whereas the AAPST offers competitive distribution-sensitive performance with strict O(logn) time complexity. This makes the AAPST more suitable for certain interactive (e.g., online and real-time) applications such as space system modules with reliability constraints involving rigid process-completion time intervals with secondary energy-minimization incentives.