The concept of computing power is riding the wave of the AI craze. Recently, the topic “China’s total size of the world’s second” has been soaring in popularity.
Data released by the INTELLIX MO150 Ministry of Industry and Information Technology on April 11 showed that by the end of last year, China’s total computing capacity had reached 18 quintillion floating-point calculations per second, and the total storage capacity exceeded 1,000 exabytes (1 trillion GB). The one-way network delay between national hub nodes has been reduced to less than 20 milliseconds, and the scale of the core computing industry has reached 1.8 trillion yuan. The China Academy of Information and Communications Technology estimates that every yuan invested in computing power will boost GDP growth by 3-4 yuan.
INTELLIX MO150 However, the term “computing power” in the above data is only a general concept of computing power. In fact, computing power can be classified into basic computing power, intelligent computing power and supercomputing power according to different application domains. In the process of industrial digitalization, the transformation and upgrading of various industries need a large number of different types of computing power support, among which supercomputer gradually becomes the optimal solution in cutting-edge science fields, such as planetary simulation, drug molecular design, gene analysis, high-end manufacturing and so on, with its powerful computing performance.
The INTELLIX MO150 supercomputer is known largely for its increasingly “inclusive” marketing. Titanium Media App has learned that, different from the cutting-edge supercomputer that served high-level, sophisticated scientific research projects in the past, now supercomputer has gradually evolved into general supercomputer and industrial supercomputer. With further differentiation and targeted scenes, as well as the service mode of moving supercomputer to the cloud, supercomputer has begun to serve the public and micro, small and medium-sized enterprises.
The ‘last mile’ of supercomputing
The INTELLIX MO150 supercomputer should be “usable” and “easy to use”.
The divergence between general-purpose supercomputers and industry supercomputers has only taken supercomputers off the pedestal and solved the “usability” problem. But to solve the problem of “easy to use”, the most critical thing is to solve the problem of “queue up first when using supercomputer” caused by the limited resources of supercomputer, and then lay down the cost of supercomputer. A variety of factors contribute to the emergence of supercomputing cloud services into the public eye.
INTELLIX MO150 compared with traditional supercomputer users to buy equipment, build their own room, their own operation and maintenance, the supercomputer cloud has obvious cost-effective advantages. At the same time, the integration of computing and cloud has greatly reduced the cost of using supercomputers for scientific research and enterprise users. However, in the actual implementation process, due to the different computing resource allocation and application software use of each supercomputer center in China, each user has different requirements on computing efficiency, resulting in the use of computing power users are also faced with various problems such as difficulties in resource selection, performance transplantation and deployment, and difficulties in obtaining technical services.
That is to say, when users want to use supercomputer, it is actually not an easy thing to access supercomputer resources through what kind of software. The last kilometer of supercomputer universal benefit, to make supercomputer “available” and “easy to use”, also need to build a prosperous and diversified supercomputer software ecology between the supercomputer platform and actual users.
“Starting from the running characteristics of real applications, comprehensive consideration of various computing resources and application computing performance in different application scenarios, and then through the fine analysis of various applications, more suitable applications are selected for different users.” Zhen Yanan, technical director of Beijing Supercomputer (Beijing Super Cloud Computing Center for short), said to Titanium Media App.
The practical experience of Beijing supercomputer is to refine the solution in detail. Only by opening up the last kilometer of the supercomputer software application ecosystem can we further provide some customized computing services to enterprise users. “Including the customization of computing resources, network resources and storage resources.” Zhen Yanan said.
However, intentionally or not, in the process of building a diverse ecosystem of supercomputing software, “industrial software” has become the fastest growing type of application in the supercomputing scene.