Using the high-level based program interface to facilitate the large scale scientific computing

Area of Science:

• Computer Science
• High-Performance Computing

Background:

• Grid and desktop grid platforms offer distributed computing resources.
• Large-scale scientific computing faces challenges in programming, middleware overhead, and data management.

Purpose of the Study:

• To investigate methods for facilitating large-scale scientific computing on grid platforms.
• To evaluate the impact of high-level programming interfaces and data anticipation migration.

Main Methods:

• Utilized the block-based Gauss Jordan algorithm as a case study for large-scale scientific computing.
• Assessed the overhead associated with high-level middleware and the effectiveness of data anticipation migration.

Main Results:

• High-level program interfaces simplify the development and execution of complex scientific applications on large-scale platforms.
• A degree of overhead is inherent with high-level interfaces, but deemed acceptable.
• Data anticipation migration significantly improves the efficiency of platforms handling big data scientific applications.

Conclusions:

• High-level interfaces are beneficial for large-scale scientific computing on grid platforms, despite minor overhead.
• Data anticipation migration is a crucial mechanism for optimizing big data scientific application performance on these platforms.