Cloud computing is dynamic in use of hardware, storage and other computer resources and is often virtualized, offering the resources as a service over the Internet. A key characteristic of cloud computing is on-demand hardware requisition and per-usage pricing.
Both network bandwidth and CPU (Central Processing Unit) processing power is limited when visualization is done locally on a regular computer or on mobile devices. This makes efficient data transfer using compact representations of models and results, as well as full hardware utilization, key to interactive visualization. The GPU (Graphic Processing Units) is very power-efficient, which is beneficial for mobile devices. Therefore, moving computations from CPU to GPU allows impressive performance on a limited power budget. Furthermore, Clouds enable users to access systems through any device connected to the Internet.
The industry is currently using multiple and different representations of the same realities when respectively simulating, modeling and visualizing. The conversions between the representations reduces quality and introduces bottlenecks in the data flow. Current visualization methods have not solved the problem of high-quality interactive visualization. However, in the Cloud, data moves fast and independent of the CPU - and visualization and data processing can rely only on the data power of the far more efficient GPU.
Application areas usually have standard representations, like corner-point grids in reservoir simulations, NURBS in CAD, element meshes in FEM, and triangle meshes in rendering. These representations store information efficiently for their particular algorithm. A key objective of CloudViz is to develop a common infrastructure to handle visualization of data originating from different sources.
The success of heterogeneous computing has strengthened the concept of heterogeneous cloud computing, and thus, one might have GPUs processing data in the cloud as well as GPUs visualizing the results in a web-client. The processing power of GPUs has not been utilized in opening up the bottleneck of data, as most of the preprocessing by standard methods is done by the CPUs. As a result, the potential of heterogeneous computing – combining CPU and GPU processing for speed – has not yet been harnessed. One reason is that software vendors hesitate to make assumptions about the users’ availability of GPUs processors. If the availability of GPU processing power can be guaranteed by virtualized resources as a service over the Internet – i.e. Cloud computing - visualization methods that rely on GPUs can be available for any connected device.
Moving the visualization of simulation results to the Cloud has another important positive side effect, in particular for industrial users working in high security infrastructures. By performing the advanced visual rendering on GPUs in the Cloud, a standard web browser, without any plug-ins installed, can be used on the client side to access the visual representation in the Cloud. Installing plug-ins in web browsers represents a security threat in many environments, and is a limiting factor for the take-up of web based visualization applications depending on various browser plugins.
Financial contributors to the project are Ceetron, SimSurgery, Statoil and the Research Council of Norway. SimSurgery's main interest is to apply improved methods for visualization in surgery simulation and training, in addition to learning how to facilitate teacher-student cooperation through an iInternet infrastructure. Statoil's interest in the project it to apply the developed methods to visualize reservoir simulation results. Ceetron aims to apply the methods of high-quality visualization making the results available independent of location. The methods will be distributed to their industrial customers as a part of their software.
The primary objective of the project is to build expertise in Norway on methods for high-quality interactive visualization processes run on a cloud-based infrastructure of heterogeneous computers.