Chasing wind and lightning, he makes data storage more reliable!

As of 2022, my country has built 21 national big data center bases. Special network virtual dedicated lines, across thousands of mountains and rivers, have realized the synchronous transmission and off-site backup of important data in my country. With the rapid increase of massive data, the storage scale is also getting larger and larger. Ensuring the reliability of computer systems in emergencies such as natural disasters is one of the core requirements of big data centers. It has risen to a national strategy and written into my country's "14th Five-Year Plan".

▲Hu Renchong was named one of the "Top Ten Young Faculty Members" of HUST in 2022

Hu Renchong, a professor from the School of Computer Science and Technology of Huazhong University of Science and Technology (hereinafter referred to as "HUST"), is such a young scholar who has been rooted in the field of computer system data reliability for a long time. In recent years, he has addressed the three major challenges faced by large-scale distributed storage systems: "high cost, many failures, and slow expansion", and has achieved a series of results around the three core scientific issues of extreme storage overhead, efficient fault repair, and efficient storage expansion, realizing the complete technical link of efficient and reliable storage coding in "theory-algorithm-system". Based on the relevant achievements, Hu Renchong won the first prize of the Hubei Science and Technology Progress Award in 2018, the National Key R&D Program Young Scientist Project in 2021, and the National Youth Talent Program Project in 2022.

Hu Ronchong summarized his research as "three fasts": make encoding faster, repair faster, and capacity expansion faster. Chasing the wind and electricity, moving faster, and making data storage more reliable, efficient, and secure are the main tasks of his research. At present, the relevant advanced technologies are monopolized by foreign countries. Hu Ronchong is rooted in the "three fasts" research, full of enthusiasm for serving the country with science and technology. He hopes to work with national enterprises such as Hikvision, Huawei, H3C, and Inspur to work together in production, learning, and research, so as to achieve my country's independent control in the field of related core technologies, and achieve overtaking on the curve and self-reliance under unfavorable situations.

Devoting myself to data storage research

In 1985, based on the successful experience of the Junior Class, the University of Science and Technology of China (hereinafter referred to as USTC) opened the "Teaching Reform Pilot Class" (also known as the "00 Class") for students with excellent college entrance examination scores, following the model of the Junior Class. The two types of students are managed by the Junior Class Management Committee, complementing each other and bringing out the best in each other. In 2001, Hu Renchong was admitted to the Junior Class of USTC with excellent grades and became a student of the 00 Class.

The elites of the junior class gathered together, and Hu Renchong, who had always been an outstanding student, deeply felt the truth that there are always people better than you, and thus established a life attitude of lifelong learning and continuous improvement. He has been interested in computers since he was a child. In the free and flexible atmosphere of the junior class, he has been studying computer-related content in a targeted manner since his freshman year, and finally chose computer science and technology as his major.

After graduating from undergraduate studies in 2005, Hu Renchong obtained the qualification for direct doctoral studies with his outstanding performance and continued to stay at USTC to study computer software and theory. "I was full of longing for scientific research at that time, and this passion has influenced me to this day. I never cared about whether it was difficult or not. I would ask questions if I didn't understand and learn if I didn't know, and I always received positive responses from my tutors." With his own efforts and the careful teachings of his tutor Professor Xu Yinlong, Hu Renchong carried out theoretical research on reliability coding in the field of computer storage. At that time, research on new types of distributed storage coding had just emerged, and one of the founders of the cutting-edge network coding technology was Professor Yang Weihao, co-director of the Network Coding Institute of the Chinese University of Hong Kong (hereinafter referred to as "CUHK"). While listening to Yang Weihao's lectures, Hu Renchong keenly perceived the development potential of distributed storage and network coding technology, and became interested in related research, which also laid the groundwork for a teacher-student relationship later.

Distributed storage is based on network technology and uses small servers to build storage pools. It is suitable for massive storage of information due to its low cost and high scalability. However, since distributed server nodes are prone to failure and data loss, how to ensure the high reliability of data has become an urgent problem to be solved. In 2009, after in-depth research, Hu Renchong's research results on the repair mechanism of distributed storage fault tolerance were published in the top journal in the field of computer communications, IEEE Journal on Selected Areas in Communications (IEEE-JSAC). This paper is one of the important papers on the combination of early network coding and distributed storage, and it laid a good theoretical foundation for Hu Renchong's subsequent research on reliability coding technology in distributed large-scale big data centers. "I also obtained a postdoctoral position at CUHK because of this, and my postdoctoral supervisor at the time was Professor Yang Weihao."

▲Hu Renchong (first from right) organizes a research group meeting

From 2010 to 2015, Hu Renchong conducted postdoctoral research at the Institute of Network Coding and the Department of Computer Science and Engineering of CUHK. While at the Institute of Network Coding, Hu Renchong became one of the first postdoctoral researchers to apply reliability coding to cloud storage systems. The relevant results were published at the Usenix Conference on File and Storage Technologies (FAST), a Class A conference in the field of international computer storage, and were recognized by the academic community. CUHK has an open and active research environment with continuous academic exchanges. Hu Renchong was further exposed to cutting-edge ideas in the academic community here, accumulating rich background and resources for his subsequent scientific research career.

"Professor Feng Dan, Dean of the School of Computer Science and Technology at HUST, is also the Director of the Key Laboratory of Information Storage Systems of the Ministry of Education. He is well-known in the field of storage at home and abroad. My research is also inspired by the work of Professor Feng Dan's team." Hu Rongchong said. In 2015, after completing his postdoctoral work, Hu Rongchong bid farewell to Hong Kong and returned to his hometown of Wuhan with a heavy scientific research achievement. He joined Professor Feng Dan's team at the School of Computer Science and Technology of HUST as an associate professor, thus starting a new scientific research journey.

Innovative coding to support multiple fields

The Key Laboratory of Information Storage Systems of the Ministry of Education is also the Optoelectronic Information Storage Research Department of the Wuhan National Research Center for Optoelectronics and the Engineering Research Center for Data Storage Systems and Technologies of the Ministry of Education. The laboratory's research on information storage technology began in 1974 and has a history of more than 40 years. There are more than 50 teachers and more than 300 doctoral and master's students. It is one of China's research units with the largest number of researchers and first-class level in the field of information storage and an important talent training base. In Professor Feng Dan's team, Hu Renchong not only received financial support and project support from all parties, but also received guidance in mature scientific research routes and other aspects, and successively undertook important scientific research tasks. Among them, the Wide-stripe Erasure Coding (WEC) series of technologies is one of his representative achievements, and it is also an important achievement that he and his team first proposed internationally.

"Hikvision and Huazhong University of Science and Technology have a close school-enterprise cooperation relationship and jointly built the 'Joint Laboratory for Massive Information Storage'. In 2018, Hikvision put forward a demand, hoping to use large-scale coding to reduce storage costs. At that time, there were only theories in this regard internationally, and only one foreign company was doing related technical practices, but the details were not disclosed." In order to solve the problems encountered by Hikvision, Hu Renchong, based on his background in network coding technology, began to explore like looking for a needle in a haystack, constantly simulating and trying to find the best coding. During this period, he fell into a misunderstanding. "At that time, the latest coding results all belonged to the field of small-scale coding, so I started from here and worked hard on simulation, but I never got the desired results."

Later, Hu Renchong had an idea and designed large-scale coding based on the original paper on network coding technology around 2012. He led the team to be the first in the world to propose a series of technical achievements of large-scale erasure codes. The paper was published again at the "File and Storage Technology Conference" (FAST), and due to its application in the computer field, the relevant achievements were also published at the "Supercomputing Conference" (SC), the top conference in the field of computer supercomputing. After applying the relevant achievements, Hikvision not only saved 20% of storage costs, but also exceeded the storage cost reduction of Microsoft Cloud and Amazon Cloud. The research work has obtained 31 patents, 19 of which have been authorized, and won the first prize of Hubei Science and Technology Progress Award in 2018. Since then, Hu Renchong has applied the ultra-low cost advantages of large-scale erasure code technology to different fields such as high-performance computing, big data centers, and all-flash clusters to solve various different needs.

Faced with the massive growth of data and the ever-increasing storage scale, it is imperative to vigorously develop China's storage industry and seize the initiative in digital economic competition. However, faced with the current situation that the main storage medium for building traditional large-scale storage systems, mechanical hard disk drives (HDD), have been monopolized by three major manufacturers in the United States and Japan (Western Digital, Seagate and Toshiba), it is extremely difficult for China to seek breakthroughs in the HDD field. At the same time, with the gradual improvement of my country's domestic complete machine and flash memory production capacity, the time has come to study the relevant core technologies of the "HDD-free" large-scale storage system and replace HDD with solid-state drives (SSD).

"We must seize this opportunity to overtake others," Hu Renchong said firmly. In his opinion, the core technologies related to the "HDD-free" large-scale storage system will be one of the major needs of future national strategies and corporate development, and the main reason hindering its large-scale application is its high cost, which is three to five times the cost of HDD. Therefore, in 2022, he applied for and was approved for the general project "Research on Large-Scale Erasure Coding Technology for Large-Scale All-Flash Clusters with Low Redundancy Costs", applying the previously proposed large-scale coding technology to SSDs, greatly reducing the cost of SSDs and providing important technical support for domestic large-scale flash memory clusters.

In the field of high-performance computing, Hu Rongchong pointed out that the development of a new generation of high-performance computing systems will greatly enhance my country's computing power and meet the strategic needs of national development. However, as the scale of parallel computing systems expands to tens of millions of cores, processor and memory failures occur frequently, and network connectivity will also be greatly reduced due to frequent failures. Therefore, the three major performances of the system, computing, storage, and transmission, will be seriously affected by "frequent failures." In order to cope with frequent failures and ensure the reliability of tens of millions of core parallel computing systems, it is necessary to solve the scientific problem of "key performance assurance under frequent failures of large-scale parallel computing systems." Therefore, he applied for and was approved to host the National Key R&D Program "Research on Reliability Coding Theory and Technology for Large-Scale Parallel Computing Systems" Young Scientist Project.

In response to the key scientific issue of "performance assurance under frequent failures", Hu Rongchong will focus on solving three key technologies through research: low-overhead fault-tolerant technology for large-scale parallel computing system node failures, reducing the performance burden of processing node failures on the system; efficient transmission technology for low-connectivity networks of large-scale parallel computing systems, improving transmission efficiency under low-connectivity networks; fault-tolerant optimization and acceleration technology based on the fault characteristics of large-scale parallel computing systems, further improving the system's fault-tolerant performance. Finally, based on the relevant results, a verifiable parallel computing system function library, simulation software and prototype system will be built, transformed and implemented to solve practical problems.

In the past five years, Hu Renchong has published more than 50 papers with unremitting efforts, including 25 papers in Computer Society Class A conferences/journals and 15 papers in Class B conferences/journals. His technology has been applied to many national enterprises such as Hikvision, Huawei, Inspur, and ZTE.

Based on demand, industry-university-research cooperation promotes transformation

Looking back on his academic career, Hu Renchong is most grateful to his mentors and academic predecessors. "My doctoral supervisor, Professor Xu Yinlong, graduated from the Department of Mathematics at Peking University. His research focuses on theory and he is very rigorous in his work. He has cultivated my rigorous and down-to-earth scientific research style. At CUHK, one of my postdoctoral supervisors, Professor Li Baiqing, focuses on system engineering, is diligent in his research field, and has a respectable craftsman spirit. He has cultivated my habit of focusing on the depth of research and the ability of engineering practice. After coming to HUST, Professor Feng Dan has greatly broadened my understanding of scientific research. Under her guidance, I have a comprehensive grasp of scientific research work and have the ability to be independent." Hu Renchong's scientific research path has gradually been put on the right track and has continued to achieve results.

In addition to scientific research, Hu Rinchong serves as the dean's assistant, responsible for scientific research and discipline construction, and assists the college in completing the work related to the construction of double first-class universities. At the same time, he is also enthusiastic about education. Since 2015, the graduate students he has guided have won the National Scholarship for Graduate Students of Huazhong University of Science and Technology many times. He also led his students to win the TECH4ALL Special Award (4/50) of the "Huawei China University Student ICT Competition" in 2021. The research results of smart music gloves can help the hearing-impaired feel the music, which was interviewed and reported by CCTV. In addition, he has participated in teaching and research projects and won the first prize of the Excellent Book Award; participated in the Ministry of Education-Alibaba Cloud Industry-University Cooperation Collaborative Education Project. Based on his outstanding performance, Hu Rinchong was named one of the "Top Ten Young Faculty and Staff of Huazhong University of Science and Technology" in 2022.

Hu Ronchong likes to read science fiction novels. While his mind is full of wild imagination, his steps are particularly firm. He said that science fiction allows for future speculation on real technology, while scientific research makes a lot of science fiction speculation become reality. "The reason why I insist on innovation along the way is that I want to change reality." The advent of the era of artificial intelligence has also made big data centers face the situation of upgrading to smart data centers. How will the reliability demand scenarios of smart data centers develop in the future? Hu Ronchong gave his own answer: "I think compared with the past, data storage must not only be safe, but also respond faster and interact with people more efficiently. At the same time, the accuracy of data storage will decline." Faced with this new application scenario, Hu Ronchong and his peers are doing all kinds of preparations, condensing new ideas for breaking through, starting from national needs, combining industry, academia and research, so that more results can be transformed and help the country achieve technological self-reliance.