A ML-based Runtime System for Executing Dataflow Graphs on Heterogeneous Processors
Subho S. Banerjee, Steve Lumetta, Zbigniew T. Kalbarczyk, and Ravishankar K. Iyer
SoCC 2018
Abstract
This paper briefly describes the design a system that addresses the challenge of scheduling distributed data-analytics workloads on heterogeneous processing fabrics (which include CPUs, GPUs, FPGAs and ASICs) in cloud-based, dynamic, multi-tenant environments. To demonstrate the capabilities of the proposed system, we use a computational genomics workflow that addresses the growing need for rapid genomic analyses in hospital environments.
Citation
@inproceedings{Banerjee2018_SOCC,
author = {Banerjee, Subho S. and Athreya, Arjun P. and Kalbarczyk, Zbigniew and Lumetta, Steven and Iyer, Ravishankar K.},
title = {A ML-based Runtime System for Executing Dataflow Graphs on Heterogeneous Processors},
booktitle = {Proceedings of the ACM Symposium on Cloud Computing},
series = {SoCC '18},
year = {2018},
isbn = {978-1-4503-6011-1},
location = {Carlsbad, CA, USA},
pages = {533--534},
numpages = {2},
url = {http://doi.acm.org/10.1145/3267809.3275474},
doi = {10.1145/3267809.3275474},
acmid = {3275474},
publisher = {ACM},
address = {New York, NY, USA},
}