The reconstruction of digital holograms on a computational grid

Abstract

Digital holography is greatly extending the range of holography's applications and moving it from the lab into the field: a single CCD or other solid-state sensor can capture any number of holograms while numerical reconstruction within a computer eliminates the need for chemical development and readily allows further processing and visualisation of the holographic image. The steady increase in sensor pixel count leads to the possibilities of larger sample volumes, while smaller-area pixels enable the practical use of digital off-axis holography. However this increase in pixel count also drives a corresponding expansion of the computational effort needed to numerically reconstruct such holograms to an extent where the reconstruction process for a single depth slice takes significantly longer than the capture process for each single hologram. Grid computing - arecent innovation in large-scale distributed processing - provides a convenient means of harnessing significant computing resources in an ad-hoc fashion that might match the field deployment of a holographic instrument. We describe here the reconstruction of digital holograms on a trans-national computational Grid with over 10 000 nodes available at over 100 sites. A simplistic scheme of deployment was found to provide no computational advantage over a single powerful workstation. Based on these experiences we suggest an improved strategy for workflow and job execution for the replay of digital holograms on a Grid.