The Cerberus Project

Final Results and Special Thanks

Final Results

Overall, the current state of the project performs under the 24fps initial goal, but this is not a critical failure. The project has been shown to operate at scale in the current implementation, as well as under optimal hardware conditions and massively scaled enterprise systems.

Once faster ethernet is utilized, the project can work under the correct speed throughput. Although this could not be installed on the current cluster due to cost and financial limitations, it was tested and proven to work at scale.

With this being said, the project can be considered a proof-of-concept success. Performance is proven to work at proper scale, and the final images are an accurate representation of the same product of a SONY proprietary software process.

Special Thanks

There are a long list of people to thank for their involvement in the project.

First off, Gurcharan S. Khanna, the primary advisor for the project was an integral part of the project. His help in adopting the UNIX mentality while developing and testing was crucial to performance. In addition, his suggestion to use BBCP for file transfers was valuable.

RIT Computer Engineering was a great help in their lending of compute nodes and Intel series servers.

David Long and Ricardo Figueroa were a tremendous help in their instruction and guidance.

RIT Research Computing for their knowledge sharing and 10GBe testing.

Additional thanks are also due to all who dealt with the loud, extremely hot, windfarm of a server rack that sat in the apartment for months.