Abstract | ||
---|---|---|
•A layered architecture to enable deep learning workloads on a traditional HPC setup.•Experiences observed during the deployment on MareNostrum, a petascale supercomputer.•Evaluation of the performance and scalability with different benchmarking workloads.•Impact of different configurations including parallelism, storage and networking alternatives.•Conclusions to guide similarly complex deployments in the future. |
Year | DOI | Venue |
---|---|---|
2019 | 10.1016/j.patrec.2019.01.020 | Pattern Recognition Letters |
Keywords | Field | DocType |
Deep Learning,Spark,DL4J,HPC,Performance,Scalability,MareNostrum | Computer vision,Computer architecture,Software deployment,Spark (mathematics),Supercomputer,Custom software,Artificial intelligence,Deep learning,Petascale computing,Mathematics,Scalability,Multitier architecture | Journal |
Volume | ISSN | Citations |
125 | 0167-8655 | 0 |
PageRank | References | Authors |
0.34 | 0 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Leonel Cruz | 1 | 0 | 1.01 |
Rubén Tous | 2 | 65 | 11.07 |
Beatriz Otero | 3 | 5 | 4.72 |