TGQR 2010Q4 Report.pdf - Teragridforum.org

2.2.9
Computer and Computation Research: eaviv—Distributed visualization (PI: Andrei
Hutanu, Louisiana State U.)
Scientific research is increasingly dependent on simulation and analysis requiring high
performance computers, distributed large-scale data, and high-speed
networks. Andrei Hutanu of LSU leads a team addressing fundamental
issues in distributed visualization design and implementation, where
network services represent a first-class resource. They built a distributed
visualization system, eaviv, and a distributed visualization application
optimized for large data, high-speed networks and interaction. eaviv
connects Spider at LSU’s Center for Computation & Technology
(CCT), Lincoln and accelerator clusters at NCSA, and a cluster at the
Laboratory of Advanced Networking Technologies (SITOLA) at
Masaryk University in the Czech Republic. The Internet2 interoperable
On-demand Network (ION) provides wide-area connectivity to support
dynamic network circuit services, and NCSA’s mass storage system
provided permanent data storage. Applications request and reserve
point-to-point circuits between sites as needed using automated control
software. eaviv supports distributed collaboration, letting multiple users
in physically distributed locations interact with the same visualization to
communicate their ideas and explore the datasets cooperatively. The team implemented a raycasting
parallel volume renderer called Pcaster as the rendering component to demonstrate the
distributed pipeline’s workflow. Compared to parallel volume renderers in existing software,
Pcaster is a purely GPU-based volume renderer and image compositor supporting high-resolution
rendering. Pcaster asynchronously couples with parallel data servers for network-streamed data
input. The team tested Pcaster with data sets up to 64 gigabytes per timestep and achieved
interactive frame rates of five to 10 frames per second on Lincoln to produce render images of
1,024 x 1,024 resolution. The eaviv system’s configurable architecture also allows it to run with a
local data server and a single renderer, making it a desktop tool for small-scale local data.
This research was published in 2010 in Scalable Computing: Practice and Experience and
Computing in Science and Engineering, and was presented at the TeraGrid 10 conference.
2.2.10
Computational Mathematics: Residual Based Turbulence Models for Large Eddy
Simulation on Unstructured Tetrahedral Meshes (PI: Arif Masud, U. Illinois)
Development of computational methods for modeling turbulent
flows is considered a formidable challenge due to the plethora of
associated spatial and temporal scales. Professor Arif Masud and
graduate student Ramon Calderer from the Department of Civil
and Environmental Engineering at the University of Illinois at
Urbana-Champaign used Abe at NCSA to develop residual-based
methods for modeling turbulence in complex fluid flows and
fluid-structure interactions. Their new codes are mathematically
consistent and robust and provide high-fidelity solutions at
reduced computational costs compared to the more traditional
Direct Numerical Simulations (DNS).
In an effort to model flow-induced vibrations in off-shore oil
platforms, the team used their new method to investigate fluidstructure
interaction and turbulence around rigid and oscillating
cylinders. Underwater currents trigger periodic vortices around
risers that are oil pipelines extending from the floating platforms
Figure 2.9. eaviv is a
visualization application with
three distributed components
connected by high-speed
networks: data, rendering, and
viewer.
Figure 2.10. Vortex shedding created by
an oscillating cylinder: the cylinder was
kept motionless and the Reynolds number
increased in order to study the vortical
structures in the cylinder's wake.
18