A LITTLE BIT CONCERNING OF PARALLEL COMPUTING

In theory, the concept of parallel computing is about the applying of multiple CPUs to a single computational problem. 

In practice parallel computing requires a lot of learning efforts which include :

• parallel computer architectures
• parallel models
• algorithms and programming
• parallel run-time environment

The main issues in parallel computing is to built efficient parallel systems, which is the large or small systems that capable of exploiting greater or smaller degree of parallelism.

Different parallel computer applications require different levels of parallel performance ; in term of : 

1. Hardware Requirement -  current industry trends that satisfy most of current application 
2. Software Requirement -  current industry trends generally do not satisfy most of scientific problems.  

The increase in computational power directly relates to the increase in number of connected processors and also interconnection networks. Achieving high performance becomes significantly more difficult with increase in communication time.

For the year 2000, performance at TFLOPS level seem to be the target. These high performance levels require thousands of interconnected CPUs and also inter-communication becomes one of the crucial factors to achieve the high performance.

In term of visualization, parallel computer developer need advanced techniques to analyze the large data sets in high performance computations. These should include :

• scaleable parallel visualization methods
• hierarchical methods for representation and visualization
• data mining techniques for feature localization
• integration with data archiving. retrieval and analysis systems
• scaleable visualization methods for performance monitoring and for debugging of large parallel application

As a conclusion, writting and debugging a parallel program is much more complicated than a sequential program. Different architectures, different programming models are suitable for different applications.