Performance Modeling of Interactive, Immersive Virtual Environments for Finite Element Simulations

Abstract

Interactive, immersive virtual environments allow ob servers to move freely about computer-generated three- dimensional objects to explore new environments. The effectiveness of these environments is dependent on the graphics used to model reality and the end-to-end lag time (i.e., the delay between a user's action and the display of the result of that action). In this paper, the authors focus on the latter issue, which has been found to be equally important as frame rate for interactive displays. In particu lar, the authors analyze the components of lag time result ing from executing a finite element simulation on a multi processor system located in Argonne, Illinois, connected via ATM to the interactive visualization display located in San Diego, California. The primary application involves the analysis of an automotive disk brake system that was demonstrated at the Supercomputing '95 conference as part of the Information Wide Area Year (I-WAY) project, which entailed the interconnection of various supercom puting centers via a high-bandwidth, limited-access ATM network. The results of this study indicate that the major components of the end-to-end lag are simulation, synchro nization, and rendering times; the use of the ATM network resulted in the network time constituting only a small fraction of the end-to-end lag time.

A Study of Speed of the Boundary Element Method as applied to the Realtime Computational Simulation of Biological Organs

“…For example, [18] mentions that a neurosurgeon takes about 0.25 seconds to perform each incremental stage of surgical cutting in practice; hence it may be all right for the simulations to take 0.25 seconds to complete in this case. Also, [44] mentions that the allowable lag time is 0.1 seconds for inexperienced users and it is 1 second for experienced users; hence the simulations can take 0.1 or 1 second in this case. References [45][46][47] mention that the just noticeable difference of the human sensory system for force perception at human fingertips is about 10 Present work also serves to make some record of the performance (like speed) that can be obtained by present day typical computing hardware (like a desktop computer, a graphics processing unit (GPU), a computer cluster) together with contemporary software (like MATLAB, MATLAB Parallel Computing Toolbox, Fortran, MPI, BLACS, ScaLAPACK).…”

“…For example, [18] mentions that a neurosurgeon takes about 0.25 seconds to perform each incremental stage of surgical cutting in practice; hence it may be all right for the simulations to take 0.25 seconds to complete in this case. Also, [44] mentions that the allowable lag time is 0.1 seconds for inexperienced users and it is 1 second for experienced users; hence the simulations can take 0.1 or 1 second in this case. References [45][46][47] mention that the just noticeable difference of the human sensory system for force perception at human fingertips is about 10…”

Performance comparison of checkpoint and recovery protocols

“…Distributed systems have been used extensively to solve long-running scientific applications such as weather prediction [1], aerodynamics simulations [2], structural finite element analysis [3], computational fluid dynamics [4] etc. Many of these applications run for hours, if not for days.…”

UbiWorld: an environment integrating virtual reality, supercomputing, and design

“…The latency of the total system (including latency from the graphics system, the tracking systems, the networks and the computation engines) cannot exceed 100 ms -1000 ms. depending on the user's experience level [14]. Taylor et al studied this phenomenon in multiple-tracked, networkconnected VR systems and offered data and r e search issues to be investigated to mitigate the latencies in the system [15].…”