Cluster Computing:

CTC CLUSTERS ASSIST WORLD LEADER IN DIGITAL PRINTING

Printing digital images using small, colored particles as ink is a multibillion dollar industry. Researchers at NexPress Solutions, Inc, a Kodak company, have developed a model for digital printing using small toner particles that allows process parameters to be changed in controlled simulations to illustrate specific cause and effect relationships important to the production of images. The model is the most extensive and complete depiction of multiparticle behavior and interactions developed in the imaging industry. The breakthrough was made in collaboration with research associates at the Cornell Theory Center (CTC) who helped adapt computer code for parallel processing and used the Center’s Windows-based high-performance computing (HPC) clusters to generate the simulations.

The original computer code was a computationally intensive serial program. For example, on a 2.8GHz dual Xeon, it would take three weeks, running around the clock, for the program to finish. CTC applied its expertise in parallel computing to the problem and ported the application to one of the clusters in its 1500+ Windows/Intel-based Velocity cluster complex. The program is presently configured to run on 20 cluster nodes and improved efficiency by a factor of 20.

"Understanding how toner particles appear where they do on the printed page is key to retaining a competitive edge in the digital printer business, yet has traditionally been more of an art than a science," said Dr. Eric Stelter, Senior Scientist with NexPress. "With the help of CTC, researchers at NexPress computationally simulated the behavior of more than ten thousand toner particles at the microscale level and at nearly nanoscale resolution. CTC ported the model to a Windows HPC cluster, parallelized it, and optimized it. We used the computing power of HPC to test hypotheses on engineering simulations in a matter of hours rather than weeks."

Direct observation is not an option in most digital printing processes: toner particles are minuscule and fully contained within equipment. Some processes are disturbed by the light that is required to observe them. The traditional product improvement approach required lengthy experimentation, constant inferences regarding the behavior of toner particles, and inherently had difficulty in establishing causal relationships because many variables were linked.

"Given that few products remain on the market for more than five years, the trend for innovation has continued to escalate," said Stelter. "NexPress turned to simulation as an efficient method for improving product performance."

The simulations performed on CTC’s clusters generate movies of the internal workings of digital printing equipment and track the motion of individual toner particles. To gain more insight into the behavior of individual particles, a technique called final state coloration was used. After each movie was generated, particles were colored depending on their position at the end of the movie. In the frame shown below, all red particles will eventually be deposited at the top of the frame (on the image that is being printed) and all blue particles will end up on the bottom of the frame. This allows particles to be tracked easily, and facilitates identification of conditions that allow particles to be rapidly deposited on the image.

For NexPress, the movies have identified subprocesses operating within existing systems that had not been considered in prior research. NexPress researchers are now able to significantly increase the speed of the printing process. This was done by using the movies to write empirical equations describing the motion of particles in the digital printers, and to identify rate-limiting factors. Further testing with actual equipment verified the insights from the simulation, which were used to optimize products in development.

An added bonus: corporate executives have found the movies to be highly effective in communicating the product research results throughout the company and to select customers.

"The relationship with the consultants was key," said Dr. Ulrich Mutze, Senior Scientist with NexPress. "CTC helped us conceive the problem in a way that could be parallelized, write the parallel code, and debug the system. We did a substantial amount of programming ourselves, but CTC offered first-hand experience that accelerated the process."

"We know, with far more accuracy than ever before, how our printers work and how to control key parameters," said Stelter. "The performance improvement has been significant, because the collaboration with CTC has allowed us to generate insights that we’ve been able to use in actual, physical devices. We’ve seen results, and anticipate that the collaboration will continue to help us gain an even greater competitive advantage. The ability to scale our application on Windows HPC clusters, rather than workstations, means we will continue to bring improved, more innovative products to market faster."

About the Cornell Theory Center

CTC is a high-performance computing and interdisciplinary research center located on the Ithaca campus of Cornell University with additional offices in Manhattan. CTC currently operates an Intel/Windows cluster complex consisting of more than 1,500 processors, in addition to Unisys ES7000 Servers. Scientific and engineering projects supported by CTC represent a vast variety of disciplines, including bioinformatics and computational biology, behavioral and social sciences, computer science, engineering, finance, geosciences, mathematics, physical sciences, and business. CTC offers assistance to institutions that are considering implementing Windows-based HPC systems and provides online and hands-on training in all aspects of Windows HPC, from systems management to code porting. For more information, visit http://www.tc.cornell.edu or http://www.ctc-hpc.com.