CISL Annual Report 2005 > Production cyberinfrastructure accomplishments

Data analysis and visualization accomplishments

In early FY2005, the Scientific Computing Division (SCD) established the Data Analysis Services Group (DASG) in recognition of the growing needs of the UCAR Data Analysis community. The main focus of the group is to provide the research community with a highly advanced computing environment tailored toward the specialized needs of interactive data post-processing, analysis, and visualization. The following are significant areas of achievement in support of the data analysis community for FY2005.

Production visualization services

In addition to providing visual computing and analysis resources, DASG also provides highly specialized visualization services to the UCAR scientfic community. Highlights for the year include:

1. Visualization of outputs from the CISL's experimental High Order Method Modeling Environment (HOMME) model.
   
   

   High Order Method Modeling Environment output.

2. Visualization of numerous high-resolution turbulence simulations from HAO and ASP.
   
   

   Stratified, decaying turbulence from Kimura, Y., Herring, J., and James, R.

Visualization Lab Upgrade Project

The SCD Visualization Lab provides a state-of-the-art facility for creating and viewing scientific visualization imagery on behalf of the UCAR scientific community. As scientific computing has moved away from costly proprietary computing systems and toward less expensive commodity components, visualization practitioners have also begun to migrate toward inexpensive Linux-based PCs, equipped with powerful but low-cost graphics cards. The VisLab Upgrade Project is an effort to transition SCD's own visual computing environment from proprietary visual supercomputers to commodity workstations.

Significant accomplishments for the lab this year include:

1. Replacement of the VisLab's stereo movie server, an aging and costly-to-maintain SGI Onyx2, with a small Linux cluster, based entirely on commodity components and OpenSource software.
2. Deployment of the Storm visualization cluster: a 64-bit commodity system offering performance superior by most all metrics to the lab's proprietary visual supercomputers.
3. Deployment and evaluation of an Intel Itanium-based SGI Prism, a shared memory, multi-pipe graphics workstation.
4. Deployment of a new image-based remote visualization service that will further enable the UCAR community to take advantage of the VisLab's visualization resources.
5. Replacement of the lab's central server with a new SGI Origin 300.

Analysis Environment Project

During FY2005, DASG embarked on several tasks aimed at learning how the needs of the analysis commumnity have evolved, identifying key areas for improvement, and deciding how to effectively build a new integrated analysis and visualization environment. The culmination of these efforts was the formalization of the Analysis Environment Project. This project represents a significant step toward the realization of DASG's mission, addressing several key components in moving us toward a high-performance analysis environment, on par with NCAR's renowned supercomputing capabilities.

The project's long-term goal is to provide an improved and unprecedented integrated environment for data analysis, visualization, and data access. A substantial strengthening of present capabilities is essential for allowing the scientific community to effectively analyze model output generated by NCAR's supercomputers. Specific objectives planned to meet this goal include:

1. Providing online, shared, secure, high-bandwidth storage with adequate capacity and performance to meet the scientific community's data analysis and post-processing needs. The target capacity should provide sufficient space for individual scientists or groups to maintain a reasonable data working set online.
2. Developing a scalable computing solution, tailored specifically to the needs of scientific data analysis and post-processing, that will meet current as well as future requirements.
3. Integrating traditional analysis and visualization resources into a single, cohesive interactive analysis environment.
4. Provide connectivity between computational resources, analysis resources, online storage and archival resources.

Significant accomplishments this year include:

1. Conducted a survey of the data analysis community to determine requirements for a new analysis server and to learn what the largest deficiencies are in the current environment.
2. Determined that while tempest, the current large data analysis machine, is aging and in need of replacement, no vendor has a current offering that provides a direct replacement.
3. Propsed a new solution for computational servers that allows us to move away from single, very large SMP machines and begin to integrate high-end visualization into our data analysis servers.
4. Conducted initial analysis of vendor computational server offerings.
5. Conducted inital analysis of available share storage technologies.
6. Developed an architectural design for a testbed of candidate servers and storage systems.