Evaluation of master-slave approaches for 3D reconstruction in electron tomography

Abstract

Electron tomography allows structure determination of complex biological specimens. The tomographic reconstruction algorithms require an extensive use of computational resources and considerable processing time to compute high resolution 3D reconstructions. High performance computing (HPC) turns out to be essential to cope with these demands. We propose and evaluate different HPC strategies based on the well-known master/slave paradigm for tomographic reconstruction. Our results demonstrate that there is an underlying problem to tackle, if the performance is to be further improved: the access to the shared file system. On the other hand, it has been shown that it is possible to find out the optimal size of the tasks distributed by the master, specially for large datasets. © 2009 Springer Berlin Heidelberg.