Abstract
Many biological problems, such finding recurring geometrical patterns in the secondary structures of protein pairs, are often solved by using parallel applications running on HPC systems that, thanks to their powerful architecture and high number of CPUs, can yield good performance. Recently cloud computing is emerging as a convenient environment to deploy certain types of parallel applications. This work examines Cross Motif Search, an application that has been successfully executed in parallel on on-premise clusters and HPC systems, and studies its porting in a cloud environment. The work uses profiling and analytical modelling to predict communication overhead. While profiling gives unreliable estimates, model-based predictions and actual data are in good match, thanks to the simple pattern of communication embedded in the application. Overall, Cross Motif Search has a viable implementation in the cloud.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call