Search for: All records

On large-scale high performance computing (HPC) systems, applications are provisioned with aggregated resources to meet their peak demands for brief periods. This results in resource underutilization because application requirements vary a lot during execution. This problem is particularly pronounced for deep learning applications that are running on leadership HPC systems with a large pool of burst buffers in the form of flash or non-volatile memory (NVM) devices. In this paper, we examine the I/O patterns of deep neural networks and reveal their critical need of loading many small samples randomly for successful training. We have designed a specialized Deep Learning File System (DLFS) that provides a thin set of APIs. Particularly, we design the metadata management of DLFS through an in-memory tree-based sample directory and its file services through the user-level SPDK protocol that can disaggregate the capabilities of NVM Express (NVMe) devices to parallel training tasks. Our experimental results show that DLFS can dramatically improve the throughput of training for deep neural networks on NVMe over Fabric, compared with the kernel-based Ext4 file system. Furthermore, DLFS achieves efficient user-level storage disaggregation with very little CPU utilization. 

Nuclear spin optical rotation (NSOR) has been investigated as a magneto‐optical effect, which holds the potential for applications, including hybrid optical‐nuclear magnetic resonance (NMR) spectroscopy and gradientless imaging. The intrinsic nature of NSOR renders its detection relatively insensitive, which has prevented it moving from a proof of concept to a method supporting chemical characterizations. In this work, the dissolution dynamic nuclear polarization technique is introduced to provide nuclear spin polarization, increasing the signal‐to‐noise ratio by several thousand times. NSOR signals of¹H and¹⁹F nuclei are observed in a single scan for diluted compounds, which has made this effect suitable for the determination of electronic transitions from a specific nucleus in a large molecule.

 

Nuclear spin optical rotation (NSOR) has been investigated as a magneto‐optical effect, which holds the potential for applications, including hybrid optical‐nuclear magnetic resonance (NMR) spectroscopy and gradientless imaging. The intrinsic nature of NSOR renders its detection relatively insensitive, which has prevented it moving from a proof of concept to a method supporting chemical characterizations. In this work, the dissolution dynamic nuclear polarization technique is introduced to provide nuclear spin polarization, increasing the signal‐to‐noise ratio by several thousand times. NSOR signals of¹H and¹⁹F nuclei are observed in a single scan for diluted compounds, which has made this effect suitable for the determination of electronic transitions from a specific nucleus in a large molecule.

 

Parallel File Systems (PFSs) are frequently deployed on leadership High Performance Computing (HPC) systems to ensure efficient I/O, persistent storage and scalable performance. Emerging Deep Learning (DL) applications incur new I/O and storage requirements to HPC systems with batched input of small random files. This mandates PFSs to have commensurate features that can meet the needs of DL applications. BeeGFS is a recently emerging PFS that has grabbed the attention of the research and industry world because of its performance, scalability and ease of use. While emphasizing a systematic performance analysis of BeeGFS, in this paper, we present the architectural and system features of BeeGFS, and perform an experimental evaluation using cutting-edge I/O, Metadata and DL application benchmarks. Particularly, we have utilized AlexNet and ResNet-50 models for the classification of ImageNet dataset using the Livermore Big Artificial Neural Network Toolkit (LBANN), and ImageNet data reader pipeline atop TensorFlow and Horovod. Through extensive performance characterization of BeeGFS, our study provides a useful documentation on how to leverage BeeGFS for the emerging DL applications. 

2D nanosheets have been widely explored as electrode materials owing to their extraordinarily high electrochemical activity and fast solid‐state diffusion. However, the scalable electrode fabrication based on this type of material usually suffers from severe performance losses due to restricted ion‐transport kinetics in a large thickness. Here, a novel strategy based on evaporation‐induced assembly to enable directional ion transport via forming vertically aligned nanosheets is reported. The orientational ordering is achieved by a rapid evaporation of mixed solvents during the electrode fabrication process. Compared with conventional drop‐cast electrodes, which exhibit a random arrangement of the nanosheets and obvious decrease of rate performance with increasing thickness, the electrode based on the vertically aligned nanosheets is able to retain the original high rate capability even at high mass loadings and electrode thickness. Combined electrochemical and structural characterization reveals the electrode composed of orientation‐controlled nanosheets to possess lower charge‐transfer resistances, leading to more complete phase transformation in the active material.