More efficient network class loading through bundling

Authors:
David Hovemeyer;William Pugh
Affiliations:
Department of Computer Science, University of Maryland;Department of Computer Science, University of Maryland
Venue:
JVM'01 Proceedings of the 2001 Symposium on JavaTM Virtual Machine Research and Technology Symposium - Volume 1
Year:
2001

Citing 6
Cited 1

Automated hoarding for mobile computers

Proceedings of the sixteenth ACM symposium on Operating systems principles
Dynamic class loading in the Java virtual machine

Proceedings of the 13th ACM SIGPLAN conference on Object-oriented programming, systems, languages, and applications
Compressing Java class files

Proceedings of the ACM SIGPLAN 1999 conference on Programming language design and implementation
Reducing transfer delay using Java class file splitting and prefetching

Proceedings of the 14th ACM SIGPLAN conference on Object-oriented programming, systems, languages, and applications
Practical experience with an application extractor for Java

Proceedings of the 14th ACM SIGPLAN conference on Object-oriented programming, systems, languages, and applications
JAZZ: an efficient compressed format for Java archive files

CASCON '98 Proceedings of the 1998 conference of the Centre for Advanced Studies on Collaborative research

A heap de/compression module for wireless Java

Proceedings of the 3rd international symposium on Principles and practice of programming in Java

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper, we describe bundling, a technique for the transfer of files over a network. The goal of bundling is to group together files that tend to be needed in the same program execution and that are loaded close together. We describe an algorithm for dividing a collection of files into bundles based on profiles of file-loading behavior. Our motivation for bundling is to improve the performance of network class loading in Java. We examine other network class loading mechanisms and discuss their performance tradeoffs. Bundling is able to combine the strengths of both on-demand strategies and archive-based strategies. We present experimental results that show how bundling can perform well in a variety of network conditions.