CNCC2024
Brief introduction of the forum:
Unraveling the underlying logic of large model training: the battle of deep integration of network and computing
Time: 13:30-17:30, October 26
Location: 3rd Floor, Summer Garden
Note: If there is any change, please refer to the final information on the official website (https://ccf.org.cn/cncc2024).
The application of large-scale model technology in all walks of life is becoming more and more common due to its unprecedented intelligence. However, this superior performance requires powerful computing and storage resources.
Although computer network technology has made great progress, in the era of large models, with the continuous expansion of model scale, the demand for data transmission and synchronization has increased sharply, so how to intelligently manage and optimize network resources, reduce communication overhead and effectively support the training and inference of large models has become a common concern of academia and industry. In addition, security vulnerabilities in large-scale data transmission and storage processes may lead to serious privacy leakage risks, which puts forward new requirements for the design of network architectures.
This seminar will focus on the challenges and opportunities faced by computer networks in the context of the "big model era". The seminar will discuss how to support efficient training and inference of large models and improve network performance and resource utilization through advanced network architecture and protocol design. In addition, the forum will focus on cybersecurity and privacy protection, and discuss innovative security solutions and privacy protection mechanisms to ensure data security and network reliability. Through these discussions, we hope to promote in-depth exchanges between academia and industry, promote continuous innovation in computer network technology, and jointly address future challenges.
Forum Agenda
| order | topic | Keynote speaker | unit |
| 1 | Research on the realization of efficient data plane for computing-network integration | Xie Gaogang | University of Chinese Academy of Sciences |
| 2 | Multi-modal large models enable 6G weak communication | Yang Kun | Nanjing University |
| 3 | Research on key technologies for computational scheduling of mixed workloads in heterogeneous computing platforms | Du Junchao | Xidian University |
| 4 | High-performance interconnection communication technology for supercomputing and intelligent computing | Dong Dezun | National University of Defense Technology |
| 5 | A preliminary study on the collaborative computing and security of device-cloud large models | Wang Zhibo | Zhejiang University |
| Speculation | Zhai Ennan | Ali | |
| Liu Dongping | AWS | ||
| Yan Xingyu | Wisdom is magnificent |
Introduction of the chairman and guests of the forum
Chair of the Forum
Zhang Yuchao
He is a tenured associate professor at Beijing University of Posts and Telecommunications, a doctoral supervisor, and the director of the Network and Communication Software Center
He is a senior member of CCF, an academic AC member of YOCSEF headquarters, a member of the Internet Special Committee, a member of the Blockchain Special Committee, and a member of the Distributed Computing and Systems Special Committee, mainly engaged in research in the direction of computer network, artificial intelligence, and network security. He has presided over and participated in national, provincial and ministerial projects such as the National Key R&D Program, the National Natural Science General Project, the Youth Project, and the Key Special Project of the Beijing Natural Science Foundation, and has published more than 60 high-level papers at home and abroad as the first/corresponding author, obtained 25 national invention patents, 6 United States/PCT patents, completed 1 English monograph and 3 series as the first author, participated in the formulation of 4 national industry standards, and was selected into the Beijing Science and Technology Rising Star Talent Program in 2023.
Co-Chair of the Forum
Cui Laizhong
Professor of Shenzhen University, Vice Dean of the School of Computer Science and Software
He is an outstanding member of CCF, a member of the Standing Committee of the Internet Special Committee, a member of the Big Data Expert Committee, a member of the Blockchain Special Committee, a distinguished professor and doctoral supervisor of the School of Computer and Software of Shenzhen University, the deputy dean of the School of Computer and Software of Shenzhen University, the executive director of the Guangdong-Hong Kong Modern Information Service Provincial and Ministerial Collaborative Innovation Center, and the assistant director of the Guangdong Laboratory of Artificial Intelligence and Digital Economy (Shenzhen). His research interests include: Internet architecture, edge computing, blockchain, distributed machine learning, and multimedia networks. He has served as an associate editor of IEEE TCC, IEEE TNSM, IEEE IoTJ and other journals. He has presided over key projects of the National Natural Science Foundation of China, national key research and development projects, outstanding youth teams of the Natural Science Foundation of Guangdong Province, and Shenzhen Excellent Youth (the first session). He has published more than 100 papers in important journals and conferences at home and abroad. He was selected as a Young Changjiang Scholar of the Ministry of Education and a Young Pearl River Scholar of Guangdong Province.
Forum Speaker
Xie Gaogang
Professor, University of Chinese Academy of Sciences
He is a senior member of CCF, a researcher, a doctoral supervisor, and a professor at the University of Chinese Academy of Sciences, mainly engaged in the research of computer network architecture and distributed computing systems. He has been funded by the National Science Fund for Distinguished Young Scholars, key research and development projects, and the relevant results have been published in academic conferences and journals such as SIGCOMM, SIGMOD, JSAC, ToN, etc., and won the second prize of the National Science and Technology Progress Award, the first prize of the Technological Invention Award of the Institute of Electronics, and other awards.
Title: Research on the Realization of Efficient Data Plane for Computing Network Integration
Abstract:Services such as large model training and AR/VR require the network to provide low-latency and high-bandwidth packet transmission and processing, and it is very important to build a flexible and efficient data plane. This report analyzes the network transmission performance bottleneck of the actual application system, and discusses the implementation mechanism of packet forwarding algorithm, transmission protocol and in-network computing to build an efficient data plane.
Yang Kun
Professor of Nanjing University, Dean of the Institute of Intelligent Networks and Communications
He is a senior member of CCF, a National Distinguished Professor of Nanjing University, the Dean of the Institute of Intelligent Networks and Communications, an academician of the European Academy of Sciences (Academia Europaea) MAE, a national high-level talent candidate, and an IEEE Fellow. Ph.D. in Electronic and Electrical Engineering, University College London (UCL), United Kingdom, and Bachelor's and Master's in Computer Science, Jilin University. His main research interests include: wireless networks and communications, the convergence of communication, computing, and perception, and AI technology-enabled networks and communications. He has published more than 500 papers in international core journals and major conferences, and more than 50 patents in many countries. He is a founding member of IEEE InterCloud and one of the six executive standing committee members, a reviewer of the WMC GSMA GLOMO Awards (the prestigious Ba Award in the industry), a member of the editorial boards of several IEEE journals (such as WCM, TVT, TNB), and an associate editor of IET Smart Cities journals.
Title: Multimodal Large Model Empowers 6G Weak Communication
Abstract:The lecture focuses on how generative AI can empower weak communication with poor channel conditions in special scenarios (the current mainstream is strong communication with high speed and large bandwidth). After a brief introduction to the evolution of 6G system, the challenges faced by the current communication network are proposed, and then the idea of a new generation of artificial intelligence AI technology represented by neural network and machine learning to help the endogenous intelligence of communication network is introduced, and semantic communication and how to use multimodal large models to empower weak communication are introduced, and then specific implementation suggestions and specific implementation technology examples and prototype system demonstrations are given.
Du Junchao
Professor of Xidian University
Outstanding member of CCF, member of the Standing Committee of the Internet Special Committee, member of the Standing Committee of the Embedded System Special Committee, leading professor and doctoral supervisor of "Huashan Scholars" of Xidian University, international talent, executive deputy director of the Blockchain Technology Application and Evaluation Engineering Research Center of the Ministry of Education, backbone of the Shaanxi Provincial Key Laboratory of Intelligent Human-Computer Interaction and Wearable Technology, head of the Shaanxi Provincial Science and Technology Innovation Team, chief scientist of the Qin Chuangyuan Scientist + Engineer Team in Shaanxi Province, executive director of ACM Xi'an Branch, and member of the Organizing Committee of ACM China Turing Conference. He won the second prize of technological invention of the Ministry of Education in 2022 and the special prize/first prize of science and technology in Shaanxi Provincial Colleges and Universities in 2021/2019. Published the national key book publishing planning textbook "ZigBee Technology Principles and Practical Practice". He has undertaken a number of key and provincial projects of the National Natural Science Foundation of China. He has published many papers in international flagship conferences such as ACM Ubicomp, ACM Mobisys, IEEE/ACM journals, etc., and won the 2017 ACM Ubicomp Outstanding Paper Award (CCF Class A).
Title: Research on Key Technologies for Computing Scheduling of Mixed Workloads in Heterogeneous Computing Platforms
Abstract:In the heterogeneous computing platform, the computing framework and scheduling technology are the key to the multi-objective computing requirements of mixed workloads such as deep learning model training, inference, and big data processing. In view of the challenges faced by heterogeneous computing platforms with heterogeneous computing resources and complex requirements for mixed loads, this report analyzes the computing frameworks and scheduling technologies at home and abroad, and introduces the team's research work on the portrait technology of heterogeneous computing resources and mixed workloads, the coupling and affinity computing technology between computing power supply and complex load demand, and the computing scheduling to ensure the complex and diverse service quality of mixed workloads.
Dong Dezun
Researcher at National University of Defense Technology
Researcher and doctoral supervisor of the School of Computer Science and Technology of the National University of Defense Technology, Changjiang Scholar of the Ministry of Education, winner of the National 100 Outstanding Doctoral Dissertation Award, winner of the National Defense Outstanding Youth Fund, winner of the Hunan Outstanding Youth Fund, the main research direction is computer architecture, high-performance and intelligent computing, parallel and distributed systems, as the deputy chief designer of the galaxy/Tianhe, he has participated in the development of domestic high-performance computer systems for a long time, and has been awarded the second-class merit of the army, and won the special prize of Hunan Provincial Teaching Achievement Award, the first prize of military scientific and technological progress, He has published more than 70 papers in A/B international journals recommended by CCF, and was selected into the 2017 China Association for Science and Technology Outstanding Scientific and Technological Paper Selection Program, and won the runner-up of the 2022 IEEE Transactions on Computers Best Paper Award. He is a member of the editorial board of Computer Engineering and Science, a distinguished member of CCF, a member of the Standing Committee of the Architecture Committee, a member of the Standing Committee of the Distributed Computing and Systems Committee, and an executive member of the High Performance Computing Committee.
Title: High-performance interconnection communication technology for supercomputing and intelligent computing
Abstract:In high-performance and intelligent computing systems, with the growth of task parallelism scale and computational model complexity, interconnection communication has become one of the key factors restricting the scalability of the system. In order to improve the scalability of parallel intelligent systems, it is necessary to think and try from different levels and dimensions, including exploring the high-bandwidth/low-latency/low-overhead design optimization of the interconnection microarchitecture within the chip, the end-to-end/full-protocol stack optimization of topology/routing/flow control/communication libraries between chips, and the collaborative optimization of application-driven algorithms/computing/communication at the system level.
Wang Zhibo
Professor of Zhejiang University and Deputy Dean of the School of Cyberspace Security
He is an outstanding member of CCF, a member of the Standing Committee of the Internet of Things Special Committee, a professor of Zhejiang University, the deputy dean of the School of Cyberspace Security, and a national excellent young man. He is currently a senior member of ACM/IEEE/Institute of Electronics, a member of the Standing Committee of the Intelligent Information Network Committee of the Society for Artificial Intelligence, and the deputy secretary-general of the Cyberspace Security Expert Committee of the Institute of Electronics. His research interests include intelligent Internet of Things, artificial intelligence security, data security and privacy protection, and he has published more than 70 CCF Class A papers, many of which have won the Best Paper Award and the Best Student Paper Award, and have been selected as the top 2% of the world's top scientists in 2023. He has presided over a number of national-level projects such as the National Excellent Youth Project, the Joint Fund Key Project, and the Science and Technology Innovation 2030-New Generation Artificial Intelligence Major Project, and his research results have served Huawei, Alibaba, Ant Financial, Hangzhou City Brain and other enterprises. He has won many awards such as the first prize of the Natural Science Award of the Institute of Electronics, the first prize of the Natural Science Award of Zhejiang Province, and the Mid-career Research Achievement Award of the IEEE Extensible Technology Committee.
Title: A Preliminary Study on Collaborative Computing and Security of Device-Cloud Large Models
Abstract:The wave of large models represented by GPT-4 has swept the world and has become a key engine to empower thousands of industries and drive the development of new quality productivity. With the continuous improvement of the intelligence of terminal devices and the privacy concerns of users when using large models, the deployment and application of large models in terminals will become a future trend. However, resource-constrained terminals are unable to support the computing requirements of large models, so it is urgent to combine cloud-side resources with collaborative computing to enable large model training and inference. This report will discuss the needs and challenges faced by the collaborative computing of device-cloud large models, and share the team's accumulation in device-cloud collaborative computing and security, as well as the attempts made on the large model.