Monitoring and Migration of a PETSc-based Parallel Application for Medical Imaging in a Grid computing PSE

In last decades, imaging techniques became central to the diagnostic process providing the medical community with a fast growing amounts of information held in images. This implies developing computational tools which allow a reliable, robust and efficient processing of data and enhanced analysis. Moreover, clinicians may have the need to explore collaborative approaches and to exchange diagnostic information from available data. A medical experiment often involves not a single approach but a set of processings that should be sometimes executed concurrently. Grid computing is becoming a cost effective emerging technology for high performance computing aggregating resources that cannot be available locally [16]. In particular, grid technologies are a promising tool to deal with current challenges in medical domains. On the other hand, employing a distributed infrastructure, where nodes may be geographically scattered all around the world and not dedicated to a specific application, is not without a price. The challenge of the grid computing paradigm derives mainly from the dynamic nature of resource requirements. In this context, in particular, reliability is a key issue and critical to the correct diagnosis. Here we are concerned with improvements and enhancements of a medical imaging grid enabled infrastructure, named MedIGrid, oriented to the transparent use of resource-intensive applications for the management, processing and visualization of biomedical images [5, 8, 6]. MedIGrid has been designed so that users can schedule reconstruction jobs needed in tomographic nuclear imaging or the denoising of ultrasound images arising in 3D echocardiography. In this paper we focus on the optimization of the software routines of MedIGrid for dynamically adapting to changes in the computational nodes. More