View-Dependent Data Prefetching for Interactive Visualization of Large-Scale 3D Scientific Data

Authors

Jin Wang, University of Nebraska-LincolnFollow

First Advisor

Hongfeng Yu

Date of this Version

11-2019

Citation

Jin Wang, "View-Dependent Data Prefetching for Interactive Visualization of Large-Scale 3D Scientific Data", MS thesis, University of Nebraska-Lincoln, 2019.

Comments

A THESIS Presented to the Faculty of The Graduate College at the University of Nebraska In Partial Fulfilment of Requirements For the Degree of Master of Science, Major: Computer Science, Under the Supervision of Professor Hongfeng Yu. Lincoln, Nebraska: November, 2019.

Copyright (c) 2019 Jin Wang

Abstract

One of the most significant challenges for today's interactive visualization is the efficient analysis and visualization of large-scale data, and I/O becomes a significant performance bottleneck. This thesis proposes a new data management policy to support interactive large-scale visual analytics. Our method can characterize user's data access patterns according to their data-dependent and view-dependent visualization operations, and leverage application knowledge to derive a novel scheme to predict data access during the interactive operations. Based on the prediction results, we develop a data replacement policy to exploit data locality and minimize data movement across multiple levels of a memory hierarchy. We evaluated our approach on machines with multiple hierarchical memory levels and compared it with state-of-the-art data replacement methods to demonstrate the effectiveness of our approach.

Adviser: Hongfeng Yu.