An experimental study on hub labeling based shortest path algorithms

Ye Li, U. Leong Hou, Man Lung Yiu, Ngai Meng Kou

Research output: Journal article publicationConference articleAcademic researchpeer-review

16 Citations (Scopus)

Abstract

Shortest path distance retrieval is a core component in many important applications. For a decade, hub labeling (HL) techniques have been considered as a practical solution with fast query response time (e.g., 1-3 orders of magnitude faster), competitive indexing time, and slightly larger storage overhead (e.g., several times larger). These techniques enhance query throughput up to hundred thousands queries per second, which is particularly helpful in large user environment. Despite the importance of HL techniques, we are not aware of any comprehensive experimental study on HL techniques. Thus it is difficult for a practitioner to adopt HL techniques for her applications. To address the above issues, we provide a comprehensive experimental study on the state-of-the-art HL technique with analysis of their efficiency, effectiveness and applicability. From insightful summary of different HL techniques, we further develop a simple yet effective HL techniques called Significant path based Hub Pushing (SHP) which greatly improves indexing time of previous techniques while retains good query performance. We also complement extensive comparisons between HL techniques and other shortest path solutions to demonstrate robustness and efficiency of HL techniques.

Original languageEnglish
Pages (from-to)445-457
Number of pages13
JournalProceedings of the VLDB Endowment
Volume11
Issue number4
DOIs
Publication statusPublished - 1 Jan 2018
Event44th International Conference on Very Large Data Bases, VLDB 2018 - Rio de Janeiro, Brazil
Duration: 27 Aug 201731 Aug 2017

ASJC Scopus subject areas

  • Computer Science (miscellaneous)
  • Computer Science(all)

Cite this