A system level view of Petascale I/O on IBM Blue Gene/P

Frings, Wolfgang; Hennecke, Michael

doi:10.1007/s00450-011-0154-4

Cited by 9 publications

(2 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In recent years, the Alpha-Go of Google has defeated the world's top Go players, which has been hailed as an important milestone in the history of AI research [1,2]. Whereas, the development of AI systems that operate like human brains is based on software algorithms rather than hardware [3][4][5]. Moreover, the execution of the software learning algorithm relies primarily on a computing system based on the traditional Von Neumann architecture, which has separated memory and processor, operating in an inefficient way and consuming much energy.…”

Section: Introductionmentioning

confidence: 99%

All-metal oxide synaptic transistor with modulatable plasticity

Yang

Chen

et al. 2019

Nanotechnology

View full text Add to dashboard Cite

The artificial neural system has attracted tremendous attention in the field of artificial intelligence due to operate mode of parallel computation which is superior to traditional Von Neumann computers in processing complex sensory data and real-time situations with extremely low power dissipation. Remarkable progress has been made in the hardware-based electric-doublelayer synaptic transistors as its modulation by ion movement is similar to biological synapse for the past few years. Unfortunately, long-term potentiation (LTP) timescale is still a big challenge in hardware-based electric-double-layer synaptic transistors which is essential to processing capacity and memory formation. Meanwhile, the effect of ion concentration on the synaptic plasticity has rarely been reported. Here, a solid state electrolyte-gated transistor using Ta 2 O 5 as dielectric layer with unique ionic composition was demonstrated and the regulation of synaptic weight was realized by changing ion concentration. Both the potentiation and depression of synaptic weight such as excitatory post-synaptic current, inhibitory response (IPSC), paired pulse facilitation as well as LTP were successfully simulated. More importantly, oxygen vacancy content was tuned for the first time to modulate synaptic plasticity by varying film thickness and gas ratio, through which the intensity and duration of memory were enhanced with appropriate vacancy concentration. It indicated that appropriate vacancy concentration avoided the effects of internal electric field induced by ion excess, leading to a long-term memory. These results reveal a promising path to improve memory capacity of artificial synapse via ion modulation.

show abstract

Section: Introductionmentioning

confidence: 99%

All-metal oxide synaptic transistor with modulatable plasticity

Yang

Chen

et al. 2019

Nanotechnology

View full text Add to dashboard Cite

show abstract

“…However, this configuration also introduces the need for additional hardware and software layers to manage the transfer of data between the compute cluster and permanent storage [1], as shown in Figure 1. These layers provide tools for parallel application developers to effectively utilize the available HPC resources, while not requiring them to be aware of the inner workings of the I/O system.…”

Section: Introductionmentioning

confidence: 99%

A Visual Network Analysis Method for Large-Scale Parallel I/O Systems

Sigovan

Muelder

et al. 2013

2013 IEEE 27th International Symposium on Parallel and Distributed Processing

View full text Add to dashboard Cite

Abstract-Parallel applications rely on I/O to load data, store end results, and protect partial results from being lost to system failure. Parallel I/O performance thus has a direct and significant impact on application performance. Because supercomputer I/O systems are large and complex, one cannot directly analyze their activity traces. While several visual or automated analysis tools for large-scale HPC log data exist, analysis research in the high-performance computing field is geared toward computation performance rather than I/O performance. Additionally, existing methods usually do not capture the network characteristics of HPC I/O systems. We present a visual analysis method for I/O trace data that takes into account the fact that HPC I/O systems can be represented as networks. We illustrate performance metrics in a way that facilitates the identification of abnormal behavior or performance problems. We demonstrate our approach on I/O traces collected from existing systems at different scales.

show abstract