STANNIS: Low-Power Acceleration of Deep Neural Network Training Using Computational Storage
This addresses power efficiency and data privacy issues for large-scale neural network training, though it is an incremental improvement over existing computational storage approaches.
The paper tackles the problem of high power consumption and data movement bottlenecks in deep neural network training by proposing a distributed in-storage training framework using computational storage devices, achieving up to 2.7x speedup and 69% energy reduction with no significant accuracy loss.
This paper proposes a framework for distributed, in-storage training of neural networks on clusters of computational storage devices. Such devices not only contain hardware accelerators but also eliminate data movement between the host and storage, resulting in both improved performance and power savings. More importantly, this in-storage processing style of training ensures that private data never leaves the storage while fully controlling the sharing of public data. Experimental results show up to 2.7x speedup and 69% reduction in energy consumption and no significant loss in accuracy.