The Persistent Buffer Tree : An I/O-efficient Index for Temporal Data

TL;DR

The paper introduces a probabilistic self-balancing persistent buffer tree data structure optimized for external memory, enabling efficient multi-version data management and range queries with I/O performance comparable to single-version buffer trees.

Contribution

It presents the first I/O-efficient persistent buffer tree supporting multi-version updates and range queries with optimal expected amortized I/O bounds.

Findings

Achieves I/O-optimal performance similar to buffer trees

Supports insertions, updates, deletions, and range queries for multiple versions

Provides probabilistic self-balancing for external memory efficiency

Abstract

In a variety of applications, we need to keep track of the development of a data set over time. For maintaining and querying this multi version data I/O-efficiently, external memory data structures are required. In this paper, we present a probabilistic self-balancing persistent data structure in external memory called the persistent buffer tree, which supports insertions, updates and deletions of data items at the present version and range queries for any version, past or present. The persistent buffer tree is I/O-optimal in the sense that the expected amortized I/O performance bounds are asymptotically the same as the deterministic amortized bounds of the (single version) buffer tree in the worst case.