SUN -- NCSA Alliance: NPAC(Syracuse) Indiana and NCSA (Illinois)

Java Grande Benchmarks and Evaluation

Name of CR Activity: Java Grande Benchmarks and Evaluation

Globally we intend to work with Sun to understand how Java based environments can be built that support Grande Applications. We will develop exemplars, general techniques and work within the Java Grande Forum to provide community input and support for the development of Java Grande environments. See http://www.npac.syr.edu/users/gcf/javagrandemar98/index.html for a more detailed discussion of Java Grande.

This project has components from three institutions. The NCSA activity is centered on a particular Grand Challenge (Nanomaterials) while that at Indiana is a broader effort to collect and evaluate a suite of Java Grande kernels. The NPAC activity is described here and there are separate documents describing the NCSA and Indiana components.

Coordinate the NCSA Alliance Java Grande Activity in a way that is synergistic with goals of international Java Grande Forum. The Syracuse activity will both monitor and interacts with other two groups but have separate work that enhances the Indiana benchmarking and NCSA Nanomaterials efforts.

1. Arrange and participate in meetings of Java Grande Forum and of Alliance partners
2. Develop High level tools to build Java Grande Nanomaterials environment. See http://www.npac.syr.edu/users/haupt/ALLIANCE/index.html
3. Develop parallel and sequential benchmarks to augment and complement those from Indiana. These include Java audio and video codecs as well as HPJava parallel codes (http://www.npac.syr.edu/users/gcf/alliance98/hpjavaallianceapril98/index.html )

The "Java Grande Community" -- application and computer science

The context is defined by nature of Alliance (high quality national multi-faceted academic collaboration centered on "Grande" computing) and by nature of Java Grande Forum (high quality academic government industry) consortium centered on Java.

1. Ongoing work in Forum.
2. Ongoing Coordination Activity
3. Benchmarks first delivered by SC98
4. Java Grande Environment for Nanomaterials by JavaOne 99 and prototype by SC98
5. Ongoing work on HPJava modified in accordance with Forum discussions

We are as above aiming deliverables at 6 month intervals aimed at SC9X and JavaOne

A greater understandinf of Java Grande with examples of its use and inproved Java Grande environments

Java Grande Support - High Performance Parallel and Distributed Computing for large scale scientific applications

What is the overall CR objective?

To provide benchmarks and analysis of Java, Java Enterprise Beans the Java3D package and Java CORBA bindings as applied to large-scale scientific applications including:

a) scientific simulation,

b) distributed visualization and tele-immersion.

c) remote access to large scale scientific data bases.

d) distributed manufacturing.

The project will collect and analyze potential benchmark applications in collaboration with Sun, Syracuse University and NCSA. Who is the targeted audience? Scientific and Engineering software designers. What is the overall context of the strategy? The context is work to support the Java Grande consortium.

A report that describes where Java can play a role in scientific and engineering applications. The focus will be on those applications that require large scale parallel execution or distributed high performance computing. The analysis will focus on the overall design suitability and detailed performance studies using Java, Enterprise Java Beans, Java RMI, and Java3D and CORBA. What are the major milestones and when do they happen? The initial milestone will be to present preliminary findings at the Java Grande forum scheduled for late 1998. The final report will be due at the end of the first year of the CR.

There is great potential to open up new markets for the application of Java technology in scientific and engineering applications. However, to make this a reality, we Java technology that can scale well to handle the computation and communication requirements of massive scientific applications. This study will provide an understanding of where the current Java technology fails and what must be done to correct it.

To develop JAVA simulation tools to form an integrated, multiplatform simulation environment, to submit and analyze simulations of microscopic systems. The analysis phase is a distributed high performance application.

1. Name of CR Activity: "Java Grande for simulations of nanomaterials"

To develop JAVA basedtools to initiate and analyze simulations, particularly related to nanomaterials. Scientific objectives of the nanomaterials application technologies team within the NCSA Alliance can be found at

http://alliance.ncsa.uiuc.edu/partners/ApplicationTechnologies/Nanomaterials.html

Develop prototypes of simulation workbench. We will build on some of the work of the HPF-Java viewer, extending to meet our needs. We will also be working with the WebFlow software from NPAC to integrate various codes together.

Anyone doing Monte-Carlo or molecular dynamics simulations of many particle systems.

The nanomaterials group is developing codes to simulate materials at different levels of description, on parallel machines such as the SGI-Origin, the IBM SP-2, NT cluster etc. We are developing tools to analyse the data and to integrate various codes together. We would like to see how Java Grande works for this.