Event Abstract Back to Event Interoperability between the CBRAIN and VIP web platforms for neuroimage analysis Tristan Glatard1, 2*, Marc-Etienne Rousseau1, Sorina Camarasu-Pop2, Pierre Rioux1, Tarek Sherif1, Natacha Beck1, Reza Adalat1 and Alan C. Evans1 1 McConnell Brain Imaging Centre, Montreal Neurological Institute, McGill University, Canada 2 University of Lyon, CNRS, INSERM, CREATIS, France Accessing substantial amounts of computing resources is required by several neuroimaging studies. CBRAIN (Sherif et al, 2014) and VIP ─ Virtual Imaging Platform (Glatard et al, 2013) ─ are two web portals offering access respectively to the Canadian and European Grid Infrastructures for neuroimage analysis. They provide services to launch and monitor experiments with state-of-the-art neuroimaging tools (e.g., FSL, Freesurfer and CIVET), and to trigger the required data movements accordingly. We are developing mechanisms to facilitate the interoperability between these portals, which would help users share and access larger data sets, access more computing resources, and access richer application catalogs. In this work, we describe our solutions to (i) harmonize authentication (ii) exchange data files (iii) share computing resources (iv) exchange applications between CBRAIN and VIP. We enabled single sign-on authentication to CBRAIN and VIP using Mozilla Persona, a secure, easy-to-implement, easy-to-use system respecting users' privacy. As a result, users can log-in to both VIP and CBRAIN using their email address only. Data sharing was enabled by developing synchronization robots between the Canadian and European infrastructures. With these robots, large data sets can be exchanged asynchronously between European and Canadian infrastructures, masking most of the transfer times. Regarding resource sharing, we are interfacing CBRAIN with the DIRAC task scheduler (Tsaregorodtsev et al, 2009) so that computing resources of the European Grid Infrastructure can be leveraged in CBRAIN. Exchanging applications is more ambitious due to the difficulty to automatically deploy neuroimaging applications on heterogeneous computing systems, and due to potential reproducibility issues resulting from such cross-system deployments (Gronenschild et al, 2012). To address these issues, we are developing an architecture based on the deployment of virtual machines on clusters, grids and clouds. Our prototype (Glatard et al, 2014) implemented in CBRAIN allows to register applications in virtual disk images, to run analyses with specific disk images, to deploy VMs uniformly and automatically on clusters and clouds, and to control the performance-cost trade-off associated to the deployment.