Over the past year, we've re-written the OME analysis system to make it more robust and efficient, and added an internationalization layer that includes a complete translation of both the user interface and the underlying semantic framework into Spanish.? ? The previous analysis system relied on a single long-lived process to orchestrate the analysis work-flow, even when the algorithm executions themselves were performed in a distributed way on several computers. When performing analyses spanning several days, this process would occasionally terminate, resulting in a loss of the analysis context, though not of the results already computed. Re-building of this context for large jobs would take up to several hours, interrupting the work-flow. Additionally, this single process constituted a choke-point, reducing the number of machines that could be used efficiently in parallel. The new distributed analysis system uses short-lived processes and the database itself to orchestrate the workflow, thus making any individual failure far less consequential. Additionally, many such """"""""workflow management"""""""" processes can operate concurrently on the same analysis task without interfering with each other. With the new system, we observed a linear increase in computational performance up to 24 independent CPUs, while the old system began to saturate with as few as 12 CPUs. We expect the new system to use even more CPUs effectively.? ? Our collaborators at CVR (Centro de Visin y Robtica, Ecuador) were instrumental in the Spanish translation of OME. Our group was primarily responsible for implementing the internationalization layer in OME such that multiple languages could be supported, and switched with a single configuration setting. Our goal was to support multiple languages all the way down to the semantic and data-model layer that underlies all of OME, and not simply to make a """"""""skin"""""""" for the user interface. This is a unique example of a system with internationalization support down to the data-model layer. The user-interface was completely translated as well, making the Spanish version of OME as fully functional as the original English version.? ? These new OME features can be downloaded from our public code repository at http://openmicroscopy.org/? ? In the coming year, we plan to test our pattern-recognition application on the OMERO platform being developed in Dr. Swedlow's lab in Dundee, Scotland. This platform has matured significantly in the previous year, and support is in place for the high-performance computing necessary for image analysis based on pattern-recognition. If these tests prove successful, we will begin to transition some of our efforts to the OMERO project, beginning with our analysis system and its pattern-recognition application.
