Why is GT a true Peer-to-Peer (P2P) network?

In the GT network, every computer can help other computers' computation, and every computer can also make use of any other computer's resources as well. Each individual in the network is an independent Peer to one another. They can take advantage of one another's shared resources. The collective result is that from any individual desktop, it seems that one has a supercomputer in front of him.

What is the difference between GT and other distributed computing startups technologies?

There are several differences. The major differences are the following:

• GT is bi-directional. GT is a true P2P network, while most of others are  really client/server-based network. This means that in GT network, any individual can perform supercomputing applications, while in others, only the server can dispatch tasks to clients to perform supercomputing - individuals are contributors only. In GT, individuals are not only contributors, they are also beneficiaries as well. In GT, an individual can perform any GT-enabled supercomputing/parallel compouting applications from any computers in the GT network.
• GT is written entirely in Java. GT runtime system is cross-platform because it is written entirely in Java. Most of other systems are  platform dependent, application dependent. It requires applications to be recompiled to target different machine architectures. GT applications are written in Java so GT application developers write once, run on all platforms.

Can it be applied to both coarse and fine PP Applications?

GT's parallelism is at the thread level of a typical operating system. 
 

What's the current status (e.g. applications running, results etc.)?

The GreenTea technology is in beta trial stage. We have two demo applications, one is the Factoring application which computes factors of arbituary larges number set (30 to 90 large numbers). The other demo application is the computer graphic Ray Tracing application which generates complex computer graphics. Both applications are very compute-intensive. In a heterogeneous network of 10 machines (PCs and Sun workstations), up to 7.6 times speedup can be achieved. In certain network configurations, more than 10 times speedup is achieved, due to the fact that some helper machines are faster than task originator machines.

Currently we are looking for applications to realize the true value of the GreenTea technology. We would greatly appreciate it to hear your ideas and thoughts about applying GT technology in any potential applications in your domain area. Your input is very much welcome!

Why is the technology named "GreenTea"?

The reason we named the technology as GreenTea is that Green Tea is the Oriental equivalent of Java. GreenTea is indeed a very powerful platform,  equivalent to the Java technology. It is our desire to share this vision with any individuals and organizations in order to realize the true potential of this great technology. Currently we are looking for people and organizations to apply this technology to their specific domains. If you think you or your colleagues and friends can apply this technology in their domain area, please email us directly and we can discuss further opportunities of cooperation. 

What are the requirements and procedures to join GreenTea?

The requirements and procedures to join GT is very simple and straightforward. Any Java-enabled networked PCs/Unix/Solaris/Linux/Machintosh machines can run GT. All of the machines running GT software form a GT network where every machine can share each other's compute resources, so in GT network every machine can start a GT-enabled parallel supercomputing application. This is why GT is a true P2P network. This is particularly convenient for scientists and engineers to run supercomputing applications from their desktop concurrently by using  idle computing resources. On the business side, the software package is available through software license. Please email us for details about GT license.

Can we use GreenTea to solve a very large time consuming Scientific and Engineering problem?

Definitely yes! A large complex problem can be divided into many smaller and easier sub-problems. You would get the most benefit if your application is written in Java. The performance of Java is getting much better now than before. I'll show you the performance study between Java and C in the next question. The basic GT application programming model is:

gt.register();

for n subtasks, gt.putSubTask();

for n subtasks, gt.getResults(subtaskID);

gt.deregister();

GT's API is very simple. The above 4 APIs are the basic APIs of the GT platform.

What's the current Java's performance in comparison with Fortran, C, C++?

Here is an article comparing performance of Java with C language:  http://www.aceshardware.com/Spades/read.php?article_id=153.

As you can see, in some cases, Java's hotspot performance engine (come with JDK1.3) achieves better performance than standard C. The reason is that Java Hotspot performance engine is able to dynamically identify the hotspot of the Java program and do dynamically linking, inlining and optimization, which C language can not do such optimizations because C is statically compiled.

By using Java, GT applications, unlike other traditional parallel applications, do not require the binary/bytecode to be residing in the remote machines physically. The code will be dynamically transmitted to the  remote machines for execution. This removes a big headache in order to run a parallel application.

Is Java ready for scientific computing?

Here is a very good article to explain Java for scientific computing.There are also translator software tools that converts Fortran to Java bytecode.
Even with legacy code, GT can still contribute its value. GT is heterogeneous in nature. It runs on all platforms as long as there is JVM(Java Virtual Machine) for that platform. Currently JVM is available from Windows, Unix, Linux, Machintosh, to mobile devices. So there is huge potential in the direction of Java. GT is futuristic in visioning and pioneering technology of tomorrow, and riding the wave of enabling technologies to achieve what others can not achieve.

What is below GT? (I know it is Java but how is the communications organized?)

GT does not make use of any existing parallel communication protocals or framework such as MPI, PVM, etc because they are more complex and
cumbumsome. For example, in order to run a traditional parallel application, the binary code must be physically residing on the remote machine. If there are 50 machines, the parallel application code must be intalled in 50 machines, or must be installed to a well-known directory which each machine knows to locate the subtask binary code. In addition, the binary code must be recompiled for a specific machine/CPU/OS architecture, e.g. for Windows machine, the binary needs to recompiled to the Window's platform; for Linux machine, it needs to be recompiled to Linux platform; for Sun Solaris machine, it needs to be recompiled to Solaris platform, and so on... In GT, the GT application code do not need to be known to remote machines because the subtask code is dynamically downloaded to remote machines. Since it is Java bytecode, any Java-enabled computing devices can contribute to help executing GT subtasks. So GT network is a heterogeneous network.

The communication among GT machines is through customized RMI (Remote Method Invocation). For example, GT applications running on the GT runtime on machine A sends subtasks to other GT machines in the GT network. The subtasks running on remote machines send results back to the GT application on machine A.

Who wrote the OS of GT? Where I may get further info about it?

Chris Xie single-handedly wrote the OS of GT, as well as porting the demo applications to GT. So he can answer any questions regarding GT. Just email him.

What programming language did you use to write GT? How big is the GT software package?

GT is written entirely in Java programming language. The total size of the current version of the GT software package, including API documents, demo source code, is about 250K in winzip format. The GT kernel without the UI code is very small memory footprint. It's about 100K in Java jar format.

Are there any installation requirements (e.g. types of platforms, resources, etc.)

Any Java-enabled networked platforms can run GT. GT has been tested on Windows and Solaris unix machines. A basic PC requirement is: Pentium II CPU 200MHz, 64M RAM, with 10/100Mbps Ethernet connection. We also found GT works on dialup Internet connections.

What is the level of availability of GT? Can we obtain a version for evaluation?

GT is available for beta trial upon signing a Non-Disclosure Agreement (NDA). Please email us for details.

Where I may get further info ?

You can visit our website at www.GreenTeaTech.com, or please email chris directly as he is the inventor of the GreenTea technology.