RTK: The Rails Toolkit

 
 

RTK is an open source API that enables applications to co-schedule CPU cores, GPUs, Memory and Network resources within multi- and many- core computer systems. We define a “Rail” as the co-scheduling of two or more of these resources. Using RTK, applications will be able to create one or more rails over which their data-intensive computation and data retrieval can be accelerated without interference from other rails or applications- hence dramatically increasing performance. RTK can be used to improve the performance of high performance computing applications, high-speed data delivery applications, as well as high-resolution graphics and video streaming.


Contacts

  1. Venkatram Vishwanath    Sungwon Nam    Luc Renambot    Jason Leigh


News

  1. RTK paper accepted and presented at IEEE e-Science 2008 (December 2008)

  2. RTK 1.0 beta available for download!! (Sept 1, 2008).


Download

  1. RTK Beta v1 (Beta v2 will be released shortly)

  2. Railperf  - Iperf like network performance benchmarking tool that uses RTK API


Documentation

  1. API Documentation

  2. Railperf Howto


Papers

  1. Vishwanath, V., Leigh, J., Shimizu, T., Nam, S., Renambot, L., Takahashi, H., Takizawa, M., Kamatani, O. The Rails Toolkit (RTK) - Towards End-System Topology-Aware High End Computing , Proceedings of the 4th IEEE International Conference on e-Science, 12/07/2008 - 12/12/2008

  2. Vishwanath, V., Shimizu, T., Takizawa, M., Obana, K., Leigh, J., Towards Terabit/s Systems: Performance Evaluation of Multi-Rail Systems, Proceedings of Supercomputing 2007 (SC07), Reno, NV 11/10/2007 - 11/16/2007

  3. Vishwanath, V., Shimizu, T., Takizawa, M., Obana, K., Leigh, J., Towards Terabit/s Systems: Performance Evaluation of Multi-Rail Systems, IEEE INFOCOM 2007, Anchorage, Alaska 05/08/2007 - 05/11/2007


Acknowledgments

  1. We would like to thank both NTT Research Labs and Sharp Laboratories of America for co-sponsoring portions of this research.