Semiconductor solutions
From smartphones to LED lamps and solar panels, semiconductors are omnipresent in today’s electronic devices. Although silicon has thus far been the darling of the electronics manufacturing industry, novel materials such as perovskites—a class of crystalline compounds with tunable conducting properties—have emerged as major contenders for the title of ‘favorite semiconductor.’
However, maximizing the function and performance of perovskites requires a thorough understanding of the arrangement of atoms within the crystal lattice. This requires solving complex mathematical equations for millions of atoms to estimate material properties such as the band gap—a measure of the minimum energy required to elevate an electron to a state where it can participate in conduction. The complexity of the problem means that researchers are increasingly relying on the powerful number-crunching capabilities of supercomputers.
At the Korea Institute of Science and Technology Information (KISTI) in South Korea, the Nurion supercomputer—ranked the 14th most powerful supercomputer in the world as of November 2019—has been used to model the nanostructure of a metal halide perovskite by performing calculations on over a million atoms. Researchers at KISTI reported good agreement between simulated and experimental bandgap values for the studied material. Importantly, findings from the simulation have resulted in ideas to overcome a bottleneck in the design of light-emitting diodes and may contribute towards the development and large-scale manufacture of next-generation optoelectronic devices.
Smart factories
With the advent of the Internet of Things (IoT), factories can now rely on a network of connected sensors to monitor production processes in real time. Market intelligence firm IDC forecasts that the manufacturing sector will account for almost US$200 billion in IoT spending in 2019, with major use cases in manufacturing operations and production asset management.
Concurrently, the amount of data generated by IoT in ‘smart factories’ is expected to grow tremendously. Without accurate and timely analyses, this data remains idle and cannot be used to enhance production workflows or inform predictive maintenance on machine parts. This is where high-performance computing (HPC) becomes essential for data mining.
In Asia, companies like Huawei and Fujitsu are actively supporting smart factories with hardware and software for computationally intensive big data analytics. For example, Huawei’s Smart Factory Solution includes scalable HPC resources in the teraFLOPS range of processing power. HPC-enabled methods can also be used to track inventory and streamline supply chains, ultimately contributing to greater efficiency and lower manufacturing costs.
Data center cooling
Despite existing in the digital realm, data has a footprint in the physical world in the form of servers and networking equipment, which are typically housed in data centers. Southeast Asia has become a hotbed for data center launches in recent years, with many global IT players setting up operations in the region. For example, Alibaba Group has established a data center in Indonesia, and Facebook is building its first Asian data center in Singapore.
A major bugbear of data center operators in the tropics is heat management. For a typical facility, 40 percent of the energy consumed goes towards cooling the hardware, translating to significant operating costs and environmental impact. Seeking better cooling strategies, Keppel Group is tapping on the Aspire 1 supercomputer at the National Supercomputing Centre Singapore to model the airflow patterns of hot and cool air in its data centers. The company then uses these findings to plan and modify the layout of its server racks for the optimal dissipation of heat and consequently, reduced requirements for cooling.
Given that data is widely regarded as the ‘new oil’ of modern industry, the rising demand for data centers looks set to continue into the foreseeable future. Supercomputers thus have a part to play in ensuring that this expansion in demand remains environmentally sustainable.
This article was first published in the print version of Supercomputing Asia, January 2020.
Click here to subscribe to Asian Scientist Magazine in print.
———
Copyright: Asian Scientist Magazine; Illustrations: Lam Oi Keat/Supercomputing Asia.
Disclaimer: This article does not necessarily reflect the views of AsianScientist or its staff.
