Research Project
Grant-in-Aid for Young Scientists (B)
We developed a MapReduce framework for GPU-based heterogeneous clusters as an instance of a future large-scale new platform with heterogeneous many core processors for big data applications. We confirmed performance scalability and overhead of our MapReduce framework by running MapReduce-based graph processing applications using 256 nodes 768GPUs of the TSUBAME2.0 supercomputer. Our results lead the foundation of softwaretechnology for next-generation extreme big data processing.
All 2013 2012 2011 Other
All Presentation (13 results) Remarks (2 results)
https://github.com/koichi626/hadoop-gpu.git