Workshop on Interoperability of DOE Visualization Centers
Institution Report

Sandia National Laboratories
Constantine "Dino" Pavlakos


a. High-level directions and priorities

Programs which are contributing to visualization efforts at Sandia include ASCI, Defense Programs, MICS, PRE (Product Realization Environment), DISCOM2, and internal core computing. In the last couple of years, ASCI has become increasingly prominent. The MICS work, however, continues to play a vital role -- it provides for more fundamental research, that has enabled development of technology which we are now leveraging for other programs, such as ASCI.

Certain themes which are central to our visualization activities include:

b. Broad views on future research activities for the next two years

Large Data Visualization -- The problem of visualizing very large data continues to be an important area of research. This is because conventional and/or commercial tools are generally inadequate for handling of very large data, necessitating development of special capabilities to address the unusual requirements of our very large-scale scientific computations. Methods which we are investigating and/or expect to incorporate to enable handling of large data include:
Among recent successes, we have demonstrated the use of our parallel visualization methods to volume render a one billion cell data set at pseudo-interactive rates.

Distributed Computing/Visualization -- Sandia has a couple of key projects in this area:

Human Computer Interfaces -- We expect to continue building on successful work already done at Sandia. In particular, work in this area has resulted in the spin-off of MUSE Technologies, collaborations with the High Performance Computing Center at Stuttgart (Germany), and a new system which incorporates a haptic device for user interface control as well as data interaction. Areas of continuing research will include:

Interactivity -- This is a performance objective which transcends our visualization activities. An implicit, if not explicit, goal of all our efforts is to provide maximal interactivity, to enable exploration, versus passive observation.

Information Visualization -- Some initial work in this area has resulted in a prototype tool for discovering relationships in large databases. Development of this tool will continue. This is an emerging area which could result in other research opportunities.

Measurements to Insight -- This phrase describes work whose nature is to extract geometric models from real world data (e.g. images, medical scans, etc.) which are then processed somehow using big computing to ultimately provide insight. An interesting application of these techniques was used recently to try to recreate the sound a dinosaur may have once made.


c. Barriers or obstacles

One important challenge for those of us who are developing state-of-the-art visualization tools is the delivery of those tools into real user environments, for production use. This is an ongoing, day-to-day struggle. At the laboratories, we are constantly being pushed by somewhat opposite forces, one that demands attention to day-to-day short term needs of application environments, while another demands that we continue to demonstrate quality-and-quantity research. Our research dollars are particularly scarce, so it is difficult to justify use of research dollars/time to "productize" our research tools. Programs like ASCI have helped alleviate this problem by providing both funding and the demand for productization of many of our MICS-developed tools. However, ASCI aside, it is important to note that productization of research tools requires significant resources which can detract from research efforts when resources are limited, which they are. This issue is equally relevant to interoperability.

An issue which we face with regard to the parallel visualization tools we are developing (which we believe are critical to solving some of our large data visualization problems) relates to portability across the range of MP/distributed architectures we are faced with. Certainly, MPI has helped a great deal. However, today's massively parallel computing systems present a variety of architectural diversities, as well as system-software, storage, and networking inconsistencies, which still present significant challenges for portable parallel tools, particularly of an interactive nature.

A related issue is that of integrating separate parallel computing software components to construct larger applications. Again, functional constraints on certain MP systems make this difficult. If integration can sometimes be difficult, general interoperability of parallel software components is even more complex.

ASCI presents one of our most difficult challenges for visualization as well as other tools. ASCI applications will require an extremely high level of confidence in results. This, in turn, demands an unprecedented level of confidence in the underlying tools.

d. Visualization Tools

Commercial(ASCI common tools):
In-House:
Other:
Data Interfaces: