On the other side of cloud computing, HPC + is bringing new ideas

The National High-performance computing academic annual meeting was founded in 2005 and has been successfully held for 13 years. The conference attracted nearly 2017 participants and 40 research institutes and manufacturers participated in the Conference. Each annual meeting will focus on the research progress and development trend of high-performance computing technology, major applications of high-performance computing, and other topics. It will provide scholars in related fields with a platform for exchange and cooperation and the release of cutting-edge scientific research achievements, it promotes the deep integration of informatization and industrialization and the development of high-performance computing in China.

On April 9, October 18-20, 2018, the 14th National High Performance Computing academic Annual Meeting (HPC China 2018) was held in Qingdao. The annual meeting was themed as "HPC +, everything is computing, highlight the continuous expansion of HPC tasks and scope. The Turing Award winner Butler Lampson once proposed the concept of "Everything is computable": through the network and increasingly powerful storage capabilities, human society can be computed; and from the 21st century, intelligent activities can also be computed.

At this annual meeting, the HPC China Top list was announced. Lenovo topped the list with 40 sets of scores and won the first place in China for the fourth time. This is also a new milestone for Lenovo following the world's top HPC rankings. Scott tease, senior director of Lenovo's data center business group (DCG) global high-performance computing and AI technology, pointed out in an interview that high-performance computing is another aspect of cloud computing, supplement the shortcomings of the cloud computing economy, and HPC itself is a technology that can be widely used. If combined with artificial intelligence, the application will be wider.

HPC application space is expanding

(Scott tease, senior director of Lenovo Data Center Business Group, Global High Performance Computing and AI technology)

IDC data predicts that the global HPC market will reach 2019 billion US dollars in 15.2 billion, with a significant growth in the Chinese market. As HPC is applied in fields such as big data and AI, gradually shifting from scientific research to commercial applications, the speed of HPC update is gradually accelerated, and the market space is broad. Scott tease said in a keynote speech in HPC China 2018: "high-performance computing is a big family that brings everyone together. This is a common undertaking of mankind ."

Why is high-performance computing a common undertaking of mankind? In addition to the importance of high-performance computing for basic scientific research, Scott tease also pointed out that high-performance computing is an important supplement to the cloud computing economy. Recently, the industry has noticed that HPC users are less active in adopting public clouds. In the past, if the public cloud was unable to obtain high-performance networking, without GPUs, and so on, but now these are not problems. Where is the problem?

Scott tease believes that, first of all, public cloud users are primarily users with low computing usage or computing system utilization. They are more willing to adopt the cloud multi-tenant approach, improve the overall utilization rate through the parallel use of fewer systems and more users. However, the computing system utilization rate of HPC users is actually very high, which can reach nearly. This alone is not consistent with the economics of the cloud.

Generally, public cloud users do not store a large amount of data on the cloud, but migrate the data from the cloud back to local storage. However, HPC users are characteristic of having massive data, if massive data is stored on the cloud and then moved back to the local device, it is very expensive and complicated. This has actually become a huge obstacle for HPC users to use the cloud. Scott tease believes that the cloud should actually be part of the HPC product, but most people now treat HPC as a local system.

The third is where data is and where computing occurs. This is true for both high-performance computing and artificial intelligence. If the data is local, Artificial Intelligence Computing occurs locally. If the data is on the cloud, the workload of Artificial Intelligence also occurs on the cloud. Therefore, the data location affects the customer's decision to select the cloud or local system.

Therefore, from the perspective of the high utilization rate, data-intensive computing tasks, and computing locations of the computing system, HPC is applied in the cloud computing, especially in the public cloud era, it will not only be eroded by the supercomputing Service of the Public cloud, but also has its own broad space. Scott tease, taking a car insurance company in the United States as an example, is able to customize a unique insurance plan for every American within ten minutes, because it uses HPC clusters. When the customer calls the auto insurance company, all the data, including age and driving habits, has been analyzed.

Lenovo's commitment to high-performance computing

Because high-performance computing has its own unique value in the cloud computing era, Lenovo attaches great importance to the development of high-performance computing. Today, under the impetus of Lenovo's "three-wave Strategy", its turnover has achieved strong double-digit growth for the second consecutive quarter, reaching $11.9 billion, an increase of 19% year-on-year, hitting a new high in the same period. Lenovo DCG's ultra-large-scale data center business saw an annual growth of nearly 68%, and its gross profit margin improved by 11 percentage points year on year, the business revenue of the Software Defined data center increased by three digits (119%) year on year ).

Lenovo DCG provides four types of computing power platforms, including the steady-state computing power platform, the sensitive computing power platform, the high-performance computing platform, and the ultra-large-scale computing platform, to meet different business characteristics and needs of enterprises. When enterprises face the load of scientific computing and AI deep learning, Lenovo, with more than a decade of profound accumulation, provides enterprises with a high-performance computing platform (HPC) as an important support. Lenovo HPC is widely used in scientific research, Meteorological Prediction, computational simulation, military research, CFD/CAE, biopharmaceutical, gene sequencing, image processing, and many other fields to promote industry changes, promote national economic development.

Scott tease stressed that Lenovo has now become the world's top high-performance computing vendor. 1/4 of the world's high-performance computing systems are Lenovo-created, and Lenovo's high-performance computing customers come from 15 countries on five continents. Lenovo HPC has three research centers in China. Beijing research centers focus on hardware, Shanghai research centers focus on software, and Taipei Research Centers focus on engineering design. "We are very proud that 17 of the world's 25 top R & D universities use Lenovo's high-performance computing systems. High-performance computing has been applied to different fields of society, such as meteorological research. Many of the top-level high-performance computing R & D teams are carried out in these universities, and we are very proud of it ." Scott tease.

Lenovo's success in high performance also comes from its investment in basic innovation and user recognition. For example, Lenovo released the fourth generation of warm water cooling solution "haishen" system, using warm water at 50 ℃ for water cooling, so that the operational efficiency of the data center is increased by 50%; previously, lenovo was the first in the industry to adopt the warm water cooling technology of 45 ℃, with maturity and leadership in the industry. Lenovo's fourth-generation warm water cooling technology, "haishen", direct-to-node machine cabinet warm water cooling solution has been applied in multiple scientific research institutions: peking University is the first super computer in China to use the Lenovo dtn (direct to node) Warm Water Cooling Technology, saving Peking University 0.6 million kWh of electricity each year; lenovo's dtn warm water cooling technology has reduced the energy consumption of devices in the Munich super Computing Center by 40%.

Scott tease himself is very enthusiastic about water cooling technology. He believes that the advantage of water cooling technology is that maintenance is very easy, and all spare parts are standardized, so it is easy to replace parts. The second advantage of this heat exchanger is that it can adapt to a variety of devices in the data center, whether for network devices or storage devices, and has universal value. Especially for the upcoming e-level computing and high-density Artificial Intelligence Computing, this is also the original intention of Lenovo's "haishen" Water-cooled Technology Development Project.

The success of Lenovo's high-performance computing is also inseparable from Lenovo's product strategy: it will not develop a product because of a unique requirement of a customer, this is because there is no universal value for other customers or the whole industry. Lenovo wants to use industry-level solutions that can be reused by users of different sizes, this includes simplifying products for large enterprises and sharing them with small and medium-sized enterprises to bring greater economic benefits.

Next stop of HPC: Artificial Intelligence

(Li Wei, director of solution marketing for Lenovo Data Center Business Group China)

One hot topic of HPC China 2018 is artificial intelligence. In particular, the topic of this conference is to promote collaborative innovation of artificial intelligence, big data, and supercomputing (ABC) to meet new opportunities for the development of today's industry.

In the AI field, Lenovo DCG's vision in the AI field is to become the most trusted AI solution provider. Through easy deployment, it provides the highest-performance AI solutions, guide the customer to implement the idea of artificial intelligence. Lenovo released its self-developed AI product family, including the new AI servers thinksystem sr670, thinksystem hg680, thinksystem hg690, and the new LICO gene edition optimized for HPC and AI.

Among them, it is worth mentioning that at the HPC China 2017 last year, Lenovo released the HPC + AI intelligent supercomputing platform LICO, which not only enhanced all functions of the previous generation of LiCo, the HPC software and hardware platform integrates various deep learning applications and mainstream open-source AI frameworks, including Google's tensorflow, Facebook's caffe, and Microsoft's mxnet. Today, the new generation of Lenovo's intelligent supercomputing platform LICO includes the brand new lico hpc OS management platform, lico-Ai platform, and LICO gene edition.

Li Wei, director of solution marketing for Lenovo Data Center Business Group China, introduced at HPC China 2018 that the LICO system is a Lenovo supercomputing platform and is now a software platform for Lenovo's intelligent computing, its advantage is that it can help users more easily use supercomputing cluster systems. For example, data scientists can easily use HPC systems using the LICO platform. This is also in line with Lenovo's product strategy, that is, the industry's universality and reproducibility.

In promoting the development of HPC to artificial intelligence, Scott tease said that after the Lenovo HPC team took over the AI work a year ago, Lenovo HPC team architects and sales have changed to AI experts, the technical knowledge reserves required in these two fields are very similar. This has also become the starting point for Lenovo to do a good job in AI, because artificial intelligence experts have already been available in sales and architecture to help customers deploy artificial intelligence systems.

Although artificial intelligence looks pretty good, its current practice is very low. Scott tease stressed that at present, there are still very few commercial artificial intelligence scenarios, and only 4% of Lenovo data center users are truly engaged in artificial intelligence business applications. There are more companies that do not know what specific business problems AI can help them solve. To better promote AI implementation, Lenovo has invested $1.2 billion to build four AI innovation centers in Morris Ville, Stuttgart, Beijing, and Taipei, provides customers with excellent infrastructure architecture, AI experts, and easy-to-use Quick Start tools. Lenovo hopes to use the Innovation Center to inspire and help Lenovo customers embrace artificial intelligence and embark on the road to artificial intelligence as soon as possible to get started with artificial intelligence.

Yuan Xiaoru, Deputy Director and Professor of the Department of intelligent science, Peking University, elaborated on the significance of supercomputing to support aI at HPC China 2018. He pointed out that the rapid development of artificial intelligence relies on the continuous enhancement of computing power, and stronger artificial intelligence will also promote the improvement of computing power, which is a process of mutual promotion. For Lenovo HPC, artificial intelligence will be the next opportunity and challenge. In particular, solving high-density Artificial Intelligence Computing will further promote the development of Lenovo HPC business. Whether it's computing human society or smart activity, Lenovo HPC has its own sky. (Wen/Ning Chuan)

On the other side of cloud computing, HPC + is bringing new ideas