The CCF Distributed Computing Conference (CCF Computility 2024) and the National Conference on Open Distributed and Parallel Computing (DPCS 2024) will be held in Changchun, Jilin Province, China from July 26 to 28, 2024, with an estimated conference size of more than 1,000 people. The theme of this conference is "Computing Network: Distributed System in the Context of New Quality Productivity", which aims to provide the most professional platform for academic discussions, technical exchanges and achievement display for practitioners related to distributed systems and computing power networks. CCF Computility 2024 has prepared 9 keynote reports by academicians and other top experts, and 15 technical forums (80 invited reports), which are definitely not to be missed!
CCF Computility 2024 | 学术盛宴，大咖云集，不容错过！
Keynote speaker of the conference
Background of the Forum
This forum focuses on the latest technological innovations in the field of computing power networked storage. This forum will gather experts, scholars, researchers, and practitioners from academia and industry to discuss the new paradigm of computing power network and showcase the latest achievements in the dynamic allocation of cloud-network-edge resources. Case studies highlight the co-optimization of high-performance compute and storage to solve data-intensive challenges. At the same time, it pays attention to green sustainability, discusses energy conservation and emission reduction strategies, and leads the transformation of computing power networks to be environmentally friendly. This forum aims to build an open communication platform, promote close cooperation between industry, academia and research circles, jointly explore the new boundaries of computing power network storage technology, and promote the high-quality development of the digital economy.
| Arrangement of the report of the Computing Network Storage Innovation Technology Forum | ||
| Chair of the Forum | Professor Guangyan Zhang and Associate Researcher Dejun Jiang | |
| Invited Presentation1 | Zhang Guangyan is the deputy director of the CCF Information Storage Technology Committee, the deputy director of the CCF Computer History Working Committee, and the tenured associate professor of Tsinghua University | Coding Storage Systems: Architecture, Theory, and Methodology |
| Invited Presentation 2 | Hua Yu is a professor at Huazhong University of Science and Technology | Distributed transaction mechanism of a large memory system |
| Invited Presentation 3 | Zhang Yiming is a professor at Xiamen Key Laboratory of Intelligent Storage and Computing | A key technology for reliable storage for computing power networks |
| Invited Presentation4 | Wu Zhongjie is a senior technical expert of Alibaba Cloud | The new high-performance distributed storage system helps the construction of computing power network |
| Invited Presentation 5 | Gu Rong is a distinguished researcher at Nanjing University | Research on elastic caching for cloud-native computing power and its open-source application |
| Invited Presentation6 | Single-boat: Architect of Huawei Cloud Setup Department | Build a disaggregated data center for large-scale inference |
| Invited Presentation 7 | Zhang Jie is an assistant professor at Peking University | Progress and outlook of near-data computing |
| Invited Presentation8 | Jiang Dejun is an associate researcher at the Institute of Computing, Chinese Academy of Sciences | QoS assurance for distributed storage systems with differentiated requirements |
| Time: July 27, 2024 afternoon Venue: Anhua Holiday Banquet Center, Changchun City, Jilin Province |
Forum Chair & Introduction
Zhang Guangyan
Tsinghua University
Zhang Guangyan is the deputy director of the CCF Information Storage Technology Committee, the deputy director of the CCF Computer History Committee, the tenured associate professor of the Department of Computer Science of Tsinghua University, and the winner of the National Science Fund for Distinguished Young Scholars. He is mainly engaged in the theoretical and methodological research of large-scale data storage and analysis, including storage systems, data compression, big data computing, AI computing systems, etc. His research has been supported by national scientific research projects, including the National Science Foundation for Distinguished Young Scholars, the National Key R&D Program, the Strategic Research and Consulting Program of the Chinese Academy of Engineering, and the 973 and 863 projects. He has published more than 60 academic papers, including more than 20 papers in top international conferences and journals in the field of computer systems, such as FAST, SOSP, USENIX ATC, EuroSys, ACM TOS, IEEE TC, IEEE TPDS, etc. As the first inventor, he has obtained more than 10 invention patents in the United States and China. The research results have been applied to the storage products of many domestic backbone enterprises, and the results have been good. The graduate students supervised by him have won the "Tsinghua University Outstanding Master's Thesis" award, "Zhong Shimo Scholarship", "Siebel Scholar" and other honors.
Jiang Dejun
Institute of Computing Technology, Chinese Academy of Sciences
Jiang Dejun, Ph.D., is an associate researcher and doctoral supervisor of the Institute of Computing, Chinese Academy of Sciences, and the director of the Data System Laboratory of the Advanced Computer Systems Research Center of the Institute of Computing, Chinese Academy of Sciences. His main research interests include storage system software and architecture, operating systems, distributed systems, etc., and he has published more than 30 papers in international conferences and journals such as ATC, TACO, MSST, ICDCS, etc. He has undertaken or participated in a number of national key R&D projects, National Youth Science Foundation projects, outstanding scientific and technological activities projects for overseas students of the Ministry of Human Resources and Social Security, and Beijing Natural Science Foundation projects. He has won the project funding of the Youth Innovation Promotion Association of the Chinese Academy of Sciences, the second prize of the Beijing Science and Technology Progress Award, and the third prize of the CCF Science and Technology Progress Award.
Presentation of the guests and the presentation
Zhang Guangyan
Tsinghua University
Zhang Guangyan is the deputy director of the CCF Information Storage Technology Committee, the deputy director of the CCF Computer History Committee, the tenured associate professor of the Department of Computer Science of Tsinghua University, and the winner of the National Science Fund for Distinguished Young Scholars. He is mainly engaged in the theoretical and methodological research of large-scale data storage and analysis, including storage systems, data compression, big data computing, AI computing systems, etc. His research has been supported by national scientific research projects, including the National Science Foundation for Distinguished Young Scholars, the National Key R&D Program, the Strategic Research and Consulting Program of the Chinese Academy of Engineering, and the 973 and 863 projects. He has published more than 60 academic papers, including more than 20 papers in top international conferences and journals in the field of computer systems, such as FAST, SOSP, USENIX ATC, EuroSys, ACM TOS, IEEE TC, IEEE TPDS, etc. As the first inventor, he has obtained more than 10 invention patents in the United States and China. The research results have been applied to the storage products of many domestic backbone enterprises, and the results have been good. The graduate students under his supervision have won the "Tsinghua University Outstanding Master's Thesis" award, "Zhong Shimo Scholarship", "Siebel Scholar" and other honors.
Title: Coding Storage Systems: Architecture, Theory, and Methodology
Report Summary: Compared to multi-copy storage, encoded storage significantly reduces the overhead of data redundancy. However, encoded storage still faces some major problems, including but not limited to: slow data recovery, low write performance on some stripes, and inconsistent performance on fast devices. In order to solve these problems, we propose a storage architecture with global resource sharing, which aims to improve the internal data read, transfer, calculation, and write speeds in the process of data recovery and data reading and writing. However, there are a number of theoretical and methodological challenges to address when implementing the above architecture, including fine-grained data layout management, large-scale recovery task scheduling, and long-tail latency optimization with median latency in mind. In this report, we will explore these issues, challenges, and initial solutions.
Huayu
Hust
Hua Yu is a professor at Huazhong University of Science and Technology, a recipient of the National Science Fund for Distinguished Young Scholars, a CCF Distinguished Member and an Outstanding Speaker. His research interests include new storage devices, high-performance storage systems, and security architectures. He has published many academic papers in OSDI, ASPLOS, MICRO, FAST, HPCA and other conferences. He has served as the program chair/vice chair at international conferences such as ICDCS 2021 and ACM APSys 2019, and as a program committee member at international conferences such as OSDI, SIGCOMM, FAST, NSDI, MICRO, ASPLOS, EuroSys, etc., and is an editorial board member of the journal ACM Transactions on Storage. The research results have won 3 provincial and ministerial science and technology awards, including the first prize of the Natural Science Award of the Ministry of Education, and the best paper awards of 4 international conferences and journals such as FAST 2023.
Title: Distributed Transaction Mechanism of Large Memory Systems
Abstract:Large memory system with large capacity, high performance, non-volatile and other characteristics is an important part of the computing network, and its architecture includes various forms such as integrated memory, interconnected memory, pooled memory, etc., and its distributed transaction processing mechanism is the key to the overall performance of the computing network. The report will comprehensively and systematically expound the distributed transaction mechanism of large memory systems that support computing power networks, and introduce relevant work progress in terms of data mode, update mechanism, network protocol, version management, etc., so as to provide ideas for the further development of large memory systems in the future.
Zhang Yiming
Xiamen Key Laboratory of Intelligent Storage and Computing
Professor Zhang Yiming is currently mainly engaged in the research of cloud computing and AI computing systems, and his achievements have been applied in key business systems such as Tianhe Supercomputer and Huawei Cloud. He is the chairman of the China Computer Systems Conference (ChinaSys), a member of the editorial board of IEEE Transactions on Computers, a member of the European Computer Systems Outstanding Doctoral Dissertation Award Committee, and the chairman of the IEEE JointCloud International Conference. He has won the second prize of the National Science and Technology Progress Award, the first prize of the Natural Science Award of Hunan Province, the CCF Excellent Paper Award, and the CCF Science and Technology Progress Excellence Award.
Title: Key Technologies for Reliable Storage for Computing Power Networks
Report Summary: Efficient and reliable data storage is the key to computing power networks. This report introduces our latest research results for two important scenarios of computing power networks, AI computing and virtualized computing. First of all, the reliability of high-bandwidth memory (HBM) is one of the important issues facing AI computing. We analyze the massive HBM fault logs in the production environment, summarize their characteristics, and propose effective prediction methods. Second, KVM (Kernel-Based Virtual Machine) is the primary VM hypervisor on Linux that follows the standard virtio framework to paravirtualize the I/O devices of customer VMs. Traditionally, KVMs have relied on QEMU to implement the backend of the virtio device family, such as virtio-blk/-net, where the collaboration between KVM (kernel space) and QEMU (user space) is key to secure and flexible storage management (e.g., migration). We have conducted research on virtualized storage technology for efficient collaboration, which has significantly improved the performance of paravirtualization.
Wu Zhongjie
Alibaba Cloud
Wu Zhongjie, known as Chudao, is a senior technical expert of Alibaba Cloud and an executive member of the Storage Professional Committee of the China Computer Federation. He has been engaged in the research and development of data storage technology for a long time, and has participated in the ZNS international standard proposal in recent years, and has done some work in flash storage and software and hardware collaborative distributed storage systems.
Title: New high-performance distributed storage system helps the construction of computing power network
Abstract:Data storage is an important part of the construction process of the computing network, and the computing network poses important technical challenges to the capacity, performance and efficiency of the storage system. It is necessary to adopt the idea of software-defined storage, improve the efficiency of distributed storage systems through software-hardware collaborative design, and meet the needs of computing power networks for data storage. This report will start from the problem analysis, put forward some methods for the design of high-performance distributed storage systems for the construction of computing power networks, and demonstrate the value that new storage systems bring to services.
Gu Rong
Nanjing University
Rong Gu, Distinguished Researcher/Ph.D. Supervisor of Nanjing University, Selected for the National High-level Young Talents Program, Winner of the DAMO Academy Young Fellow Award (2023), his main research direction is cloud computing and big data systems, distributed storage management systems, and he has published more than 60 research papers in the field, including USENIX ATC, EuroSys, VLDB, KDD, ICDE, WWW, VLDBJ, IEEE TPDS, TKDE, TON, etc. He has presided over more than 10 projects such as the National Natural Science Foundation of China, the National Key R&D Program, the Special Funding Project of the China Postdoctoral Science Foundation, and the Enterprise Innovation Fund projects of Sinopec, CRRC, China Mobile, Huawei, Alibaba, ZTE, Tencent, ByteDance, etc., and won the first prize of Jiangsu Science and Technology Award, IEEE TCSC Award for Excellence (Early Career Researcher), and IEEE He is the Chairman of the Fluid Open Source Community of the Cloud Native Computing Foundation.
Title: Research on elastic caching for cloud-native computing power and its open source application
Abstract: With the rise of AI technology represented by large models, more and more data-intensive applications (AI model training, big data query, etc.) are running on cloud-native platforms with cost-saving, flexible computing power orchestration, and convenient system operation and maintenance. However, the architectural characteristics of cloud-native survival-computing separation and resource elasticity pose great challenges to traditional caching technologies. This report introduces an intelligent elastic caching technology, including core methods such as lightweight and precise cache capacity adjustment, collaborative scheduling of cache computing power, and efficient transmission of cache replicas. Furthermore, we will share the application results of the Fluid open source project based on the above technology (which has been selected by the Cloud Native Computing Foundation) and its industry leaders.
Single boat
Huawei
Shan Zhou, Architect of Huawei Cloud Architecture Department, graduated from the University of California, San Diego with a Ph.D. degree. His research focuses on improving the cost-effectiveness of data center infrastructure, including large model inference systems, decoupled memory, and large-scale distributed storage systems. At HUAWEI CLOUD, he led cloud storage hardware acceleration and serverless large model inference projects. He has published 20+ papers in top academic conferences, and his research has won the best papers in OSDI 2018, SYSTOR 2019, and FPAG 2024 Runner Up.
Title: Building a Separate Data Center for Large-Scale Inference
Report Summary: LLM-based large model inference is becoming one of the most important workloads in data centers. Large model inference presents a variety of characteristics, which puts forward very high new requirements for the underlying infrastructure of data centers. This report focuses on how to improve the cost-effectiveness of large model inference for large-scale deployment. The report first describes the current data center inference deployment architecture, analyzes the mismatch between the model and the hardware from the model structure, and summarizes the three walls of Model Serving, namely the memory wall, the scheduling wall, and the elastic wall, and then explains the industry's solutions to these bottlenecks. The report will also discuss the deployment patterns and challenges of Agent Serving.
Zhang Jie
Peking university
Jie Zhang, Ph.D., Assistant Professor, Doctoral Supervisor, and Distinguished Researcher, School of Computer Science, Peking University, was selected into the Overseas Youth Program of the National High-level Talent Program, and won the Honorary Scholar of the Intel China Academic Talent Program and the ACM SIGCSE Rising Star Award. He has been engaged in the research and design of storage systems and special processors for a long time, and is committed to solving the needs of high-performance storage systems in the era of big data and artificial intelligence from the level of computer architecture, and breaking through the bottleneck of data migration and the limitation of memory wall under the von Neumann architecture. He has published more than 50 papers in international conferences and journals, including ISCA (CCF-A, three papers), OSDI (CCF-A), HPCA (CCF-A, seven papers), MICRO (CCF-A, two papers), ASPLOS (CCF-A), FAST (CCF-A), ATC (CCF-A, two papers), and Eurosys (CCF-A).
Title: Progress and Prospects of Near-Data Computing
Report Abstract: With the advent of the era of big data, new applications such as artificial intelligence, graph computing, and big data have put forward higher requirements for the computing power and storage capacity of server clusters. However, the traditional von Neumann architecture and supporting system software have the natural disadvantage of high data migration overhead, which cannot meet the actual needs of new applications. Today's memory and storage systems are undergoing a major technological shift. Based on this advancement, researchers need to rethink and redesign the existing system organization and hardware architecture. This report focuses on our research progress in the field of data computing, and our proposed solutions can effectively reduce the overhead of a large number of software stacks, optimize computer architecture, and eliminate the physical limitations of traditional hardware.
Jiang Dejun
Institute of Computing Technology, Chinese Academy of Sciences
Jiang Dejun, Ph.D., is an associate researcher and doctoral supervisor of the Institute of Computing, Chinese Academy of Sciences, and the director of the Data System Laboratory of the Advanced Computer Systems Research Center of the Institute of Computing, Chinese Academy of Sciences. His main research interests include storage system software and architecture, operating systems, distributed systems, etc., and he has published more than 30 papers in international conferences and journals such as ATC, TACO, MSST, ICDCS, etc. He has undertaken or participated in a number of national key R&D projects, National Youth Science Foundation projects, outstanding scientific and technological activities projects for overseas students of the Ministry of Human Resources and Social Security, and Beijing Natural Science Foundation projects. He has won the project funding of the Youth Innovation Promotion Association of the Chinese Academy of Sciences, the second prize of the Beijing Science and Technology Progress Award, and the third prize of the CCF Science and Technology Progress Award.
Title: QoS Assurance for Distributed Storage Systems for Differentiated Requirements
Report Summary: In the computing power network scenario, latency-sensitive tenants with different QoS requirements and bandwidth-oriented tenants share the underlying distributed storage system. Due to interference in resource sharing, tenant SLO is difficult to guarantee and resource utilization is low. The back-end storage architecture of distributed storage systems is different, and the bottleneck points when facing the above problems are also different. This report explores the above issues and challenges, analyzes the bottlenecks of different back-end storage architectures, and introduces QoS assurance technologies for SEDA architecture and RTC architecture, aiming to maximize system resource utilization while ensuring tenant SLO.
How to register for the conference
1. Registration Criteria
| Ticket type: | Participant Identity | 6.1~7.28 |
| Conference registration fee | Professional Member of CCF | ￥2700 |
| CCF Student Member | ￥1700 | |
| Non-Member Professionals | ￥3200 | |
| Non-member students | ￥2200 |
2. How to register
| Conference registration QR code | QR code for the homepage of the conference |
Note: The conference is paid through the CCF conference management system, and the refund shall comply with the "CCF Regulations on the Refund of Conference Registration Fees".
For the specific arrangements of the above forums, please pay attention to the official website of the conference. In addition to the keynote speeches and this forum, CCF Computility 2024 will also organize 14 unique thematic forums, each of which will be chaired by top experts to bring you the most cutting-edge academic discussions and technical exchanges.
At present, the preparatory work of the conference is nearing the end, and scholars in the field of distributed computing are welcome to come to Changchun to witness and promote technological innovation.