High Performance Parallel Java with JavaParty

Abstract

To achieve better performance with Java applications, computers can be interconnected with fast networks to form a cluster making available multiple Java Virtual Machines. Unfortunately, Java does not provide an elegant, easy to use mechanism for parallel programming on clusters. JavaParty transparently adds remote objects to Java while avoiding the disadvantages of programming with remote method invocation (RMI) and many disadvantages of the message-passing approach in general. This thesis presents a performance analysis of a cluster running a Java benchmark using JavaParty. It reveals quantitative performance measurements showing a decrease in application execution time by adding more machines. In addition, this thesis presents a method to increase the performance of the cluster network in the presence of network congestion using Quality of Service (QoS).

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jun 01, 2008
Accession Number
ADA483466

Entities

People

  • Samuel Nassar

Organizations

  • Naval Postgraduate School

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Computer Networks
  • Computer Programming
  • Computer Programs
  • Computers
  • Congestion
  • Digital Communications
  • Hash Tables
  • High Level Languages
  • High Performance Computing
  • Language
  • Measurement
  • Network Protocols
  • Operating Systems
  • Programming Languages
  • Standards
  • Transport Protocols
  • Virtual Machines

Fields of Study

  • Computer science

Readers

  • Computer Networking
  • Database Systems and Applications
  • Systems Analysis and Design