In the book “The Innovator’s Dilemma”, Harvard Business School professor Clayton Christensen argues that technology is divided into disruptive and continual, and that disruptive technology often changes the landscape of an entire industry because it works in a new way or for one. Brand new groups solve pressing problems.

Today, the data storage market is once again facing a change: on the one hand, smart applications based on cloud computing, big data and artificial intelligence are blooming in various industries, bringing unprecedented data storage needs and challenges; The disruptive new technologies represented by the NVMe protocol and SCM media are accelerating the reshaping of the storage architecture, interconnection, software stack and algorithms, trying to meet various new requirements for data storage.

Therefore, the most urgent challenge in the data storage industry, academia and research community is: how to make the scientific research direction go well in the “first mile” and scientific research results to get through the “last mile”, that is, the scientific research direction is closer to the new needs of the future market, and the scientific research results are more Well play its technical value in the market.

Recently, Inspur and Huazhong University of Science and Technology announced the joint establishment of the “New Storage Joint Laboratory”, opening up new ideas for the integration of production, education and research in the field of domestic data storage. The two parties will conduct in-depth exploration of forward-looking insights and engineering in the data storage industry.

Why the “New Storage Lab”

The field of data storage has always been a prominent representative of the integration of industry, academia and research. From various underlying storage media, to individual technologies such as data deduplication and data compression algorithms, to the whole system of all-flash arrays, there are examples of industry-university-research cooperation. . Looking at China, as China’s data storage industry has entered the fast lane of development in recent years, the emergence of a large number of new storage demands has also brought new opportunities for the integration of production, education and research in domestic data storage.

This directly contributed to Inspur Storage and Huazhong University of Science and Technology coming together.

Huazhong University of Science and Technology, as one of the first world-class universities in the country, has a solid research foundation in the field of data storage. It was the first university in China to establish a national professional laboratory in the field of storage in the early 1990s, and has become a veritable talent in China’s storage industry. The cradle is also the witness of China’s storage industry from scratch and from weak to strong. As a representative of new storage, Inspur Storage has achieved rapid progress in recent years in terms of products, technologies, market share and status, and It is at the forefront of the industry in the research and exploration of new storage.

Therefore, the hand in hand between the two parties can be described as a perfect match, giving a huge space for data storage industry, academia and research. In fact, the most critical factor that brings Inspur Storage and Huazhong University together is the huge challenge brought by new storage.

Feng Dan, dean of the School of Computer Science at Huazhong University of Science and Technology, said bluntly, “The fundamental challenge of new storage is to keep up with the needs of new applications and new scenarios.”

Looking back at the development history of data storage, every technological innovation will bring huge changes and impacts to the market. Professor Feng Dan’s point of view revealed the importance of the “first mile” of current data storage scientific research innovation, that is, when the demand for data storage changes rapidly, technological innovation needs to be closely integrated with new applications, new scenarios, and new needs. And not behind closed doors.

In fact, the challenges brought by new applications and new scenarios to data performance, capacity, reliability, and data management have already begun to emerge. Taking the AI ​​applications that are blooming everywhere as an example, one of the obvious trends of AI models is huge quantification. The GPT-3 model released by OpenAI last year has a parameter volume of 175 billion and a training data set of 570 GB. Inspur released in September this year. The 1.0 huge model has a higher level of parameter scale and training data set. The parameter scale has reached an astonishing 245.7 billion, and the training data set has exceeded 5000GB. posed great challenges.

Another example is that the self-driving function has now become the focus of all car companies to participate in market competition. One thousand L4-level self-driving cars can generate up to 100 PB of data every day. Car companies need to continuously collect massive amounts of data for automatic driving algorithms. Training, and due to regulatory requirements, all data must be stored for at least 30 years, and the data storage challenges brought by the new scenarios of car companies have never been encountered before.

There are many, many new demands like these are constantly emerging. The industry-university-research community of data storage needs to closely integrate new demands, explore new architectures, new algorithms, and use new storage to cope with the great changes brought about by smart applications and digital technologies.

“The focus is to make the technology transform more quickly.” said Li Hui, general manager of Inspur’s storage product line.

Laboratory research covers the whole field of data storage

Before announcing the establishment of the joint laboratory, Inspur and Huazhong University of Science and Technology had accumulated a good foundation in the subject cooperation. Although there are many researches on these subjects, they are more focused on the cooperation of local technology and basic commonality.

The “New Storage Joint Laboratory” will cover the entire field of data storage, from the system architecture, storage interconnection, underlying media, software algorithms and other fields, and will steadily advance and deepen the cooperation between the two parties in multiple stages.

In fact, Huazhong University of Science and Technology is at the forefront of basic technology research, new technology directions, flash memory media, algorithms, protocols, etc. There is also a deep understanding of the needs, and the joint laboratory can be targeted and give full play to their respective advantages, which is expected to bring more surprises in the future.

Taking the architecture as an example, the storage architecture is facing another reshaping after years of development. The data storage medium revolution brought by flash memory, the vigorous development of many protocols such as NVMe, the accelerated popularization of multi-core processors, and many new How to better integrate the technology into the storage system requires the industry to make breakthroughs in the storage architecture.

“Like many smart applications are large-scale applications, it is a great challenge to the internal interconnection and sharing, performance, and reliability of the architecture. The architecture of data storage requires innovation in theoretical models and breakthroughs in specific implementation. “Li Hui introduced.

For another example, new media represented by flash memory have brought great changes to data storage. New media such as flash memory have brought excellent performance, but the number of stacking layers is getting higher and higher, and the process technology is becoming more and more complex, which makes flash memory face the challenge of durability. To address the challenges in aspects such as performance, reliability, etc., how to maximize the strengths and avoid weaknesses, and better utilize the value of flash memory in the storage system, has become a challenge that the industry needs to solve urgently.

Professor Feng Dan said bluntly: “In the past ten years, new storage media have developed rapidly, and each medium has its own characteristics. How to play its advantages in suitable scenarios cannot be separated from the research and development of scientific research institutions in colleges and universities. Many new technologies in storage are at the forefront of the world. How to apply these technical research results to scenarios and promote the improvement of the overall capabilities of data storage products is still a long way to go. The joint laboratory is now open to everyone. A good head.”

“Industry-university-research” community: highlighting the value of technology

In fact, in the field of data storage, there is no shortage of successful cases of “industry-university-research” integration. Take the data deduplication technology that is widely used in the field of data storage as an example. It started out as a set of algorithms developed by Li Kai, a Chinese professor of Princeton University, in the laboratory. It has been successful and has successfully transformed the field of data backup and data management, producing great value.

In China, the “industry-university-research” integration of data storage is more urgent. For many years, the biggest problem that has plagued China’s “industry-university-research” is to get through the “last mile”, how to better generate value in the market with the scientific research results in the laboratory. In recent years, the “last mile” problem of data storage industry-university-research has become more and more prominent.

On the one hand, scientific research forces like Huazhong University continue to make breakthroughs, walk at the forefront of data storage exploration, and have accumulated rich scientific research results; at the same time, new storage leaders like Inspur, which is growing rapidly in the market, For the ever-changing new demands for data storage, in addition to its own research and development system, it is eager to work with scientific research institutions to accelerate the testing and implementation of new technologies in the market, so as to seize the rare market transformation period and achieve further success.

Therefore, the “industry-university-research” of data storage must open up new ideas in terms of mechanism, research direction and achievement transformation. According to the introduction, one of the important missions of the joint laboratory is to bring innovations and breakthroughs in the field of scientific research to the development of storage products, apply them practically in engineering, and implement the technology into products, which will eventually reach users and the industry. Professor Feng Dan said, “Scientific research results must be undertaken by enterprises in order to truly play their role and value, otherwise it will be a sleeping article.”

“The focus of the laboratory is to make innovative technologies can be transformed faster.” Li Hui said when talking about the laboratory’s goals.

In the author’s opinion, the “new storage joint laboratory” jointly built by Inspur and Huazhong University of Science and Technology is an inevitable part of the development of China’s storage industry. After more than 20 years of exploration and progress, China’s storage industry is at a critical moment of “breaking the cocoon into a butterfly”. Scientific research forces and scientific research talents are increasingly abundant, and data storage manufacturers are also emerging in the market. The industry-university-research community such as the “Storage Joint Laboratory”, integrating multi-party innovation forces, will be of great benefit to the future development of the entire industry.