SYSTEMS

By Faith Singer-Villalobos, TACC: Researchers can now create compelling visualizations of massive datasets from the convenience of a local desktop or laptop-- Remote visualization enables researchers to interactively analyze very large datasets by using powerful, high-end visualization systems located at advanced computing centers. This paradigm shift is made possible through increases in network bandwidth and the development of specialized software for remote visualization. This capability removes the limitations faced by researchers accustomed to having to use local visualization systems with less computational power, memory, and storage capacity. In research supported by NASA and the Office of Naval Research, Dr. Eric Fahrenthold and graduate student Robert Rabb of UT Austin's Department of Mechanical Engineering are developing new computational models of the impact response of high strength fabric materials. One application of this research is the simulation of orbital debris impact effects on manned spacecraft. Debris shields composed of aluminum and high strength fabrics are used to protect habitable modules of the International Space Station. Computer simulation allows for the study of hypervelocity impact effects at conditions which cannot be duplicated in the laboratory, and thereby assists in the development of improved spacecraft protection systems. “The system now in place at TACC offers both parallel rendering on a high performance graphics system and remote access in real-time to the visualization results,” says Dr. Fahrenthold. “This combination is a very valuable resource to allow for rapid and continuous evaluation of 3D simulation results, a critical component of code development and validation work.” "Visualization is a vital part of the scientific discovery process, and TACC wants to make it trivially easy for every researcher who works with us to use remote visualization," says Gregory S. Johnson, manager of the Visualization and Data Analysis (VDA) group at TACC. "We've given a great deal of attention to the area of remote visualization of scientific datasets. Our goal is to deliver to the desktop the ability to simply and efficiently explore and analyze large-scale and complex 3-D data." Think of the equivalent of "Google Maps" for visualization, Johnson continues. "What's needed in the area of remote visualization is a comparable system that stresses interactivity and ease of use. Scientific visualization can, to some degree, be simplified by automating parts of the process that are commonly troublesome. We can provide this capability to the user 'from a distance' eliminating the need to install software. Moreover, the user is no longer constrained by the capabilities of locally available graphics hardware," Johnson concludes. In the traditional visualization model, the researcher conducts a computation on a remote high-performance computing system and then transfers all of the data back to a local machine to perform the visualization work. This data transfer is increasingly problematic due to large data sizes and security concerns. In addition, a researcher's local machine may not have the graphics performance necessary to render complex images. Add the researcher’s need for modern-day collaboration with colleagues in different geographic regions and it is easy to see why this model is rapidly becoming outdated. Remote visualization provides a solution to overcome these obstacles. With this new model, the datasets are rendered remotely on a high-performance visualization engine. Only the resulting images are transmitted to the researcher’s desktop or laptop, minimizing the computational resources required locally. TACC and Sun Microsystems Since 2003, TACC has worked closely with Sun Microsystems to extend the process of scientific discovery through advanced visualization. The cornerstone of TACC's remote visualization capabilities is Maverick, a one-of-a-kind Sun high performance remote visualization system. This powerful computer is a balanced system with a considerable amount of shared memory, graphics capabilities, and number-crunching horsepower. Maverick provides the powerful analysis and rendering capabilities that users need for remote visualization of large and time-critical problems. Maverick Facts Maverick has 16 times the graphics processing ability of most modern desktop computers. Maverick has 64 dual core processors, half a terabyte of memory and more than 14 terabytes of storage, allowing users to visualize the most challenging datasets. This combination of speed and graphics capability makes Maverick ideal for projects that require digesting massive amounts of data and finding relevant relationships that are then rendered into precise images. These projects include mission-critical jobs that have great societal impact like emergency response management, weather prediction, and earthquake simulation. "Collaborating with TACC has had significant influence on the product development process," says Linda Fellingham, manager of graphics and advanced visualization at Sun Microsystems. "This close collaboration very early in Sun's development process has allowed us to incorporate valuable feedback into our products. We'’ve been able to optimize features for processing and visualizing large scientific datasets and for providing remote access and collaboration capabilities to create this shared visualization resource for The University of Texas at Austin and TeraGrid communities." "In addition, Maverick has also provided an excellent testbed for Sun's DARPA High Productivity Computing Systems visualization subsystem investigations," Fellingham says. Currently, TACC and Sun are jointly developing a software system to enable interactive remote and collaborative visualization. In addition, TACC is developing a remote visualization tool that employs a knowledge-based system to partially automate (and grossly simplify) the visualization process. Following is a description of both projects: Virtual Network Computing (VNC) Approach Dr. Christopher Gilpin, director of the electron microscope facility at the UT Southwestern Medical Center at Dallas, is using electron tomography to visualize high resolution 3D biological structures. Tomograms typically have been viewed either by stepping through z slices of the volume or by making models of structures by hand tracing objects in every slice. With the help of parallel computing at TACC, Dr. Gilpin is using software contouring to produce 3D views of data sets. One such example is a project examining the structure of synapses in the brain in collaboration with Dr. Xinran Liu at UT Southwestern. Pre and post synaptic membranes are being viewed in a unique and detailed manner. "We're in the middle of a very productive phase of collaboration with TACC to develop software tools for visualizing biological images," says Dr. Gilpin "The combination of our imaging expertise with TACC's skills at implementing high performance computing is a powerful resource for our community." The Sun Visual Grid software combines Sun Grid Engine, VirtualGL, and VNC to provide management and application-transparent use of Maverick's many graphics resources. The software will allow users to log into Maverick and use the graphical resources through any web browser equipped with Java. In addition, multiple users can interact with the same desktop environment, thus enabling collaborative remote visualization. TACC currently provides this software to all users as a stand-alone capability. The next step is to integrate it into the TACC User Portal to make it even simpler to use. Users will log into the TACC User Portal (https://portal.tacc.utexas.edu/portal.html) and click on a link that takes them to their desktops on Maverick. The TeraGrid User Portal, which officially launches in May 2006, will also support this style of remote visualization. Customized Remote Visualization System In addition to TACC’s collaboration with Sun, the center is independently developing a custom remote visualization system that will be available to users through the web. This system will leverage the graphics resource management capabilities developed by Sun while adding a clean, efficient interface to remote visualization resources that simplifies the tasks of data management, analysis, and visualization for researchers in a wide range of fields. One of the primary problems with existing visualization software packages is that they require users to fit their data into formats understood by those specific applications. Furthermore, users often do not know which visualization algorithm is best suited for a particular dataset and must experiment with additional software packages. A system that can help users overcome these difficulties will increase their productivity. TACC VDA Specialists David Guzman and Greg P. Johnson, along with student intern Peru Shanmugam, are developing a system that queries users for characteristics about their data. Based on this input, the system determines the set of visualization techniques best suited to the dataset and scientific domain. The system then presents thumbnail images illustrating each technique to the user. At that point, the user can select a thumbnail, and the corresponding technique will be used to visualize the data. An early prototype is in the works and TACC intends to roll out the software in November at Supercomputing 2006.