The Programming Model Of An Enhanced P2HP-2 And Protein Structure Prediction

In the area of bioinformatics, people have widely accepted the natural idea, getting a structure of protein by stimulation computation, which requires a huge mount of computation ability. On the basis of analyzing the domestic and international network computing research, especially the high performance computing problems in bioinformatics, we developed P2P based High Performanced Computing Platform (P2HP) platform based on peer to peer (P2P) structured network, which not only adopts the mechanism of separation between the application system and core software, but also provides a convenient user developmental kit and supports various distributed applications, and support the reciprocal high-performance computing application with arborescent task picture.The programming model is the key subsystem of P2HP. In terms with programming model a one-sided message passing programming model for P2HP (OMP), I have proposed the programming model Communion Message Programming Model (CMP) of P2HP-2, which consists of software development kit (SDK). And the data flow is code and data of the single task. According to the application programming interface that CMP offer, the distribution only means cutting programming sources into small pieces instead of each code sentence. With the guidance of CMP, the whole project is divided into several procedure module, and each work node performs own part of different tasks, which communicate with each other via message through network.The Threading alignment algorithm, which attempts to match unknown amino acid array and various three-dimensional structure that has already existed, and evaluates the score of the arrays folded into suitable degree of that kind of structure, is a kind of P2HP based distributed protein tertiary structure prediction. Moreover, the distributed algorithm combines the characteristic of CMP and identity of Threading alignment itself, Divide computational process into leaf, crotch and root. Synchronously several parallel tactics on foundation of different degrees or leafs are proposed, which improved the harmony between different nodes.Finally, a performance evaluation by the simulative test and instant test is given. We have chosen 20 target sequences and 20 templates from PDB. While taking sequence and template as fixed value, if the leaf is 20 and degree is 5,the Speedup of P2HP is 2.5 when there are 4 volunteers in the P2HP. When more volunteers join, the speedup of platform can increase constantly. For example, when the worker nodes number rise to 15, Speedup is more than 8.91. The test reveals that the P2HP support the reciprocal high-performance computing application with arborescent task picture high-efficiently, and the programming model has high efficiency and can be expanded.