Article Preview
TopIntroduction
A whole set of tools and software for dMRI in Brain Imaging Field, allows the processing and visualization of dMRI data. Some of them provide interactive and intuitive interfaces for visualization, but with a minimal support of simple non-intensive processing.
On the other hand, some are toolboxes with different types of processing that can be combined to produces the expected result. They can be deployed in a computational grid to reduce execution time. Nevertheless, they require a huge effort from the user and use low-level command lines.
In our research works, we have sought how to develop an architecture that can provide an easy to use interactive interface, while providing access to both intensive and non-intensive computational capabilities. For such purpose, we have investigated current Cloud and Grid architectures to develop a platform to reach our objectives.
Grid Computing has long been the main paradigm to provide access to a significant set of computational resources. These resources are distributed across different organizations and they are federated mainly for research purposes.
Grid Portals that allow intuitive access to grids through a web interface, have kept the same access model. Usually, the user interacts through a basic web form, where he specifies the processing to be launched, the input files and some parameters.
While some scientific research have brought some rendering capabilities, inspired by the Web 2.0 trend (For example maps), advanced visualization inside a grid portal is nearly absent. Also, as all computations are considered intensives, they are all executed in the remote grid infrastructure. There is no means to execute non-intensive computations inside the grid portal.
From the Cloud perspective, some recent SaaS services like Google Docs have provided some interactivity and visualization capabilities. Their main objective, is to provide to the user a way to interact with its data hosted into the provider Cloud.
For companies such as Google, these services provide an interesting opportunity for data-mining into more diverse user's data like text documents and spreadsheets for better advertisement targeting.
For such reasons, the SaaS model is mostly data-centric and the computational axis is not a top priority. But with the recent development of browser technologies, we can now provide access to advanced computational, 2D/3D rendering and local data storage capabilities.
These capabilities hadn't the fair amount of attention they deserve, and we think that if we exploit them appropriately we can enhance the user experience and provide new kinds of Software as a Service.
In this paper, we present a new Hybrid Architecture combining Software as a Service and Computational Grids. The main objective is to simplify the work of clinicians and researchers by providing tools to aid diagnosis, with access to advanced computational capabilities, while remaining highly intuitive and interactive.
The Software as a Service part provides the required intuitive interface and manages local non-intensive computations at the client level. The Grid part manages the intensive computations that couldn't be executed on the client.
This helps us to provide a single Service that can combine different features of dMRI Software, while still being easy to use.
Although, the service we are developing called CloudMRI, targets the dMRI in Brain Imaging Field, it is based on a generic architecture we are developing called Acigna-G (Benmerar and Oulebsir-Boumghar, 2011).
In another word, the proposed architecture can be applied in any field that requires both non-intensive and intensive computation. But, we restrict ourselves in our paper to the dMRI in Brain Imaging Field.
The rest of the paper is structured as follows: In Section 2, we introduce the different relevant concepts to dMRI Brain in general. In particular, we review some of the current dMRI based Brain Tractography Algorithms and related Software.
In Section 3, we review the current state of Grid Portals, and Software as a Service Architecture. Also, we show how some of the current dMRI Brain Software related issues can be resolved with a Hybrid SaaS-Grid Architecture.
In Section 4, we review the key browser technologies relevant to the discussion of our architecture. In fact, a great part of the current work on the architecture is related to the Browser environment.
In Section 5, we present the generic proposed architecture called Acigna-G. We also discuss certain components of the resulting CloudMRI Service, derived from this generic architecture.
In Section 6, we provide a comparative study with other related works, at different levels of the architecture.