Parallel I/O Implementation on the Multiple Sequence Alignment Software ClustalW-MPI
From SNIC Documentation
|Name||Parallel I/O Implementation on the Multiple Sequence Alignment Software ClustalW-MPI|
|Description||Parallel I/O Implementation on the Multiple Sequence Alignment Software ClustalW-MPI|
This project aims for optimizing the ClustalW software, which is the multiple sequence alignment code. NSC participates in this project as the PRACE collaborator, with the aim of designing and implementing parallel I/O for sequence alignment softwares.
This project is a user support task under PRACE 2IP WP7.1, whose project ID is 2010PA1155.
Implementation procedure is as follows.
- Analysis on Code Structure
- Design of parallel I/O interface for sequence alignment input
- Implementation to the ClustalW Software
The code basically adopts the master-slave structure, so that the deployment of parallel I/O which is the collective operation necessitates the change on the structure of the code. NSC designed and experimented the parallel I/O strategy for sequence alignment softwares, by adopting the sub-group communicator. NSC endeavoured to optimize I/O-related variables for the best performance of sequence data input, and applied the designed implementation to the ClustalW software.
The employed scheme provides up to 6.8 times speed-up for sequence data input in case of using 8192 cores under BlueGene/Q system. Experiments provide the guideline to set various parameters for best I/O performances. Details are to be accessible under PRACE White Paper archive.
|Soon-Heum Ko (NSC)||NSC||Application expert||Computational fluid dynamics|