Study of Grid Computing for Next Generation Computing

Dari PaloDozen

Tittle Study of Grid Computing for Next Generation Computing
Principla Investigator Hermawan K. Dipojono
Researcher Eko Mursito Budi
Year 2007

Extended Abstract

This research aims to find the virtual computing system to support the rural Next Generation Computing (r-NGC). The first stage was to study the well known middlewares for cluster and grid computing, including:

  • Mosix (, a program that provides the services for cluster computing environment, based on MPI (Message Passing Interface). It has been installed on 8 clustered computers at the computational material laboratory TF ITB.
  • Globus (, is a leading program for grid computing, based on Java RMI (Remote Method Invocation).

About 20 computers has been configured as grid computing at the computing facility TF ITB. The study showed that Mosix and globus are really powerfull for parallel computing, especially for scientific applications where the users would write their own program using either Java, C or Fortran langguages. However, both are not helpfull for common applications such as word processors, spreadheet, nor web browser. Unfortunatelly, that common applications are the most used ones in r-NGC.

Therefore, this research has been considering other alternatives from the available open source software technology. The focus is on using X-Windows, especially to enhance the LTSP (Linux Terminal Server Project) system. A simpler, yet flexible solution has been designed, comprises of 3+1 layers as follow:

  • The Graphical User Interface (GUI) Layer, works on the local computer to handle user interaction.
  • The Virtual Application Server (VAS), for transparently executing various applications on the local as well as remote computers.
  • The Virtual Data Server (VDS), providing access to the certain user’s data for the runing applications regardless of the location of the application.
  • The Virtual Session & Authentication Server (VSAS) allows seamless user authentication across the client and the servers.

An extended study has been conducted to implement the design by using open source software. The proposed implementation would be:

  • Use the widely available open source technology, including: X-Windows, SSH, SSHFS, and LDAP.
  • Light and flexible, thus it can run as fat cient, thick client, as well as thin-client system.
  • Allows server location transparency, by using multicast discovery service.
  • Able to optimize the performance of the system using load scheduling, proposedly using Simmetric Innitiative Distribution method.

The implementation of this design will be conducted on the next research.

Kontributor: Mursito