Jianbo Dong
Research Subjects
Jianbo Dong has a total of 122 co-authors. Here are the top co-authors by number of shared publications:
- Binzhang Fu
Alibaba Cloud, Beijing, China
5 shared publications - Dennis Cai
Alibaba Cloud, Beijing, China
4 shared publications - Ennan Zhai
Alibaba Group, Hangzhou, China
4 shared publications - Jiamin Cao
Alibaba Group, Beijing, China
4 shared publications - Jiaqi Gao
Alibaba Cloud, Beijing, China
4 shared publications - Qingxu Li
Alibaba Cloud, Beijing, China
3 shared publications
Latest Publications
- Open Access
- 11 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 368-395
EPIC: Abstraction and Polymorphism of In-Network Collectives on Ethernet
- 11 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 368-395
Open AccessEPIC: Abstraction and Polymorphism of In-Network Collectives on Ethernet
- Open Access
- 11 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 550-563
Balancing and Beyond: Communication-Centric Optimizations in Expert Parallelism
- 11 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 550-563
Open AccessBalancing and Beyond: Communication-Centric Optimizations in Expert Parallelism
- Open Access
- 26 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
C2C-Explorer: An Exploration Framework for Chip-to-Chip Interconnect Architectures in LLM Cloud Computing Systems
The scaling-up of large language models (LLMs) necessitates computing systems to have multi-processor-chip architectures, elevating the importance of chip-to-chip (C2C) communication. However, designing efficient C2C hardware architectures for LLM workloads faces three key challenges: generating realistic LLM-specific C2C traffic, accurately simulating hardware-level communication at scale, and efficiently exploring the exponentially large C2C design space. We propose C2C-Explorer, an adaptive Bayesian DSE framework that integrates a LLM-workload-driven traffic generator, a scalable interconnect simulator (switch/full-mesh, up to 512 chips), and a metric-guided evaluator into a workload-to-hardware optimization pipeline, enabling systematic C2C architectural co-design under realistic LLM workloads. Validated against FPGA-based C2C prototypes, the C2C simulator achieves 2.46–8.23% end-to-end timing error across diverse traffic patterns. Its hybrid cycle & event model further accelerates large-scale simulation by up to 7.8× over a pure cycle-accurate baseline. Applied to a 32-XPU DeepSeek-R1-671B inference workload, C2C-Explorer identifies configurations that improve goodput by 44.1% and reduce memory by 98.4%. C2C-Explorer is open-source and available at https://github.com/Selinaee/C2C-Explorer.
- 26 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Open AccessC2C-Explorer: An Exploration Framework for Chip-to-Chip Interconnect Architectures in LLM Cloud Computing Systems
The scaling-up of large language models (LLMs) necessitates computing systems to have multi-processor-chip architectures, elevating the importance of chip-to-chip (C2C) communication. However, designing efficient C2C hardware architectures for LLM workloads faces three key challenges: generating realistic LLM-specific C2C traffic, accurately simulating hardware-level communication at scale, and efficiently exploring the exponentially large C2C design space. We propose C2C-Explorer, an adaptive Bayesian DSE framework that integrates a LLM-workload-driven traffic generator, a scalable interconnect simulator (switch/full-mesh, up to 512 chips), and a metric-guided evaluator into a workload-to-hardware optimization pipeline, enabling systematic C2C architectural co-design under realistic LLM workloads. Validated against FPGA-based C2C prototypes, the C2C simulator achieves 2.46–8.23% end-to-end timing error across diverse traffic patterns. Its hybrid cycle & event model further accelerates large-scale simulation by up to 7.8× over a pure cycle-accurate baseline. Applied to a 32-XPU DeepSeek-R1-671B inference workload, C2C-Explorer identifies configurations that improve goodput by 44.1% and reduce memory by 98.4%. C2C-Explorer is open-source and available at https://github.com/Selinaee/C2C-Explorer.
- Open Access
- 27 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 645-662
SyCCL: Exploiting Symmetry for Efficient Collective Communication Scheduling
- 27 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 645-662
Open AccessSyCCL: Exploiting Symmetry for Efficient Collective Communication Scheduling
- Open Access
- 27 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 453-466
Alibaba Stellar: A New Generation RDMA Network for Cloud AI
- 27 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 453-466
Open AccessAlibaba Stellar: A New Generation RDMA Network for Cloud AI
- Top Cited PapersOpen Access
- 4 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-15
Crux: GPU-Efficient Communication Scheduling for Deep Learning Training
Deep learning training (DLT), e.g., large language model (LLM) training, has become one of the most important services in multitenant cloud computing. By deeply studying in-production DLT jobs, we observed that communication contention among different DLT jobs seriously influences the overall GPU computation utilization, resulting in the low efficiency of the training cluster. In this paper, we present Crux, a communication scheduler that aims to maximize GPU computation utilization by mitigating the communication contention among DLT jobs. Maximizing GPU computation utilization for DLT, nevertheless, is NP-Complete; thus, we formulate and prove a novel theorem to approach this goal by GPU intensity-aware communication scheduling. Then, we propose an approach that prioritizes the DLT flows with high GPU computation intensity, reducing potential communication contention. Our 96-GPU testbed experiments show that Crux improves 8.3% to 14.8% GPU computation utilization. The large-scale production trace-based simulation further shows that Crux increases GPU computation utilization by up to 23% compared with alternatives including Sincronia, TACCL, and CASSINI.
- 4 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-15
Top Cited PapersOpen AccessCrux: GPU-Efficient Communication Scheduling for Deep Learning Training
Deep learning training (DLT), e.g., large language model (LLM) training, has become one of the most important services in multitenant cloud computing. By deeply studying in-production DLT jobs, we observed that communication contention among different DLT jobs seriously influences the overall GPU computation utilization, resulting in the low efficiency of the training cluster. In this paper, we present Crux, a communication scheduler that aims to maximize GPU computation utilization by mitigating the communication contention among DLT jobs. Maximizing GPU computation utilization for DLT, nevertheless, is NP-Complete; thus, we formulate and prove a novel theorem to approach this goal by GPU intensity-aware communication scheduling. Then, we propose an approach that prioritizes the DLT flows with high GPU computation intensity, reducing potential communication contention. Our 96-GPU testbed experiments show that Crux improves 8.3% to 14.8% GPU computation utilization. The large-scale production trace-based simulation further shows that Crux increases GPU computation utilization by up to 23% compared with alternatives including Sincronia, TACCL, and CASSINI.
- 22 June 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Micro
- Vol. 41  (5) , 85-92
ACCL: Architecting Highly Scalable Distributed Training Systems With Highly Efficient Collective Communication Library
Distributed systems have been widely adopted for Deep Neural Networks (DNNs) model training. However, the scalability of distributed training systems is largely bounded by the communication cost. We design a highly-efficient collective communication library, namely Alibaba Collective Communication Library(ACCL), to build distributed training systems with linear scalability. ACCL provides optimized algorithms to fully make use of heterogeneous interconnects simultaneously. And the experimental results show significant performance improvement.decoupling the algorithm and communication operations. By leveraging ACCL, we propose a hybrid algorithm to fully make use of the intra- and inter-node interconnects simultaneously. The experimental results show that the communication performance of ACCL is about 3x over NCCL. And the end-to-end performance of applications can be improved by 12.6% to 25.2%, with respect to the ratio of communication time.
- 22 June 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Micro
- Vol. 41  (5) , 85-92
ACCL: Architecting Highly Scalable Distributed Training Systems With Highly Efficient Collective Communication Library
Distributed systems have been widely adopted for Deep Neural Networks (DNNs) model training. However, the scalability of distributed training systems is largely bounded by the communication cost. We design a highly-efficient collective communication library, namely Alibaba Collective Communication Library(ACCL), to build distributed training systems with linear scalability. ACCL provides optimized algorithms to fully make use of heterogeneous interconnects simultaneously. And the experimental results show significant performance improvement.decoupling the algorithm and communication operations. By leveraging ACCL, we propose a hybrid algorithm to fully make use of the intra- and inter-node interconnects simultaneously. The experimental results show that the communication performance of ACCL is about 3x over NCCL. And the end-to-end performance of applications can be improved by 12.6% to 25.2%, with respect to the ratio of communication time.
- Open Access
- 16 May 2016
- Published by Association for Computing Machinery (ACM)
- p. 145-153
P-Socket
Data centers require efficient, low-cost, flexible interconnects to manage the rapidly growing internal traffic generated by an increasingly diverse set of applications. To meet these requirements, data center networks are increasingly employing alternatives such as RapidIO, Freedom, and PCIe, which require fewer physical devices and/or have simpler protocols than more traditional interconnects. These networks offer raw high performance communication capabilities, but simply using them for conventional TCP/IP-based communication fails to realize the potential performance of the physical network. Here we analyze causes for this performance loss for the TCP/IP protocol over one such fabric, PCIe, and we explore a hardware/software solution that mitigates overheads and exploits PCIe's advanced features. The result is P-Socket, an efficient library that enables legacy socket applications to run without modification. Our experiments show that P-Socket achieves an end-to-end latency of 1.2μs and effective bandwidth of up to 2.87GB/s (out of a theoretical peak of 3.05GB/s).
- 16 May 2016
- Published by Association for Computing Machinery (ACM)
- p. 145-153
Open AccessP-Socket
Data centers require efficient, low-cost, flexible interconnects to manage the rapidly growing internal traffic generated by an increasingly diverse set of applications. To meet these requirements, data center networks are increasingly employing alternatives such as RapidIO, Freedom, and PCIe, which require fewer physical devices and/or have simpler protocols than more traditional interconnects. These networks offer raw high performance communication capabilities, but simply using them for conventional TCP/IP-based communication fails to realize the potential performance of the physical network. Here we analyze causes for this performance loss for the TCP/IP protocol over one such fabric, PCIe, and we explore a hardware/software solution that mitigates overheads and exploits PCIe's advanced features. The result is P-Socket, an efficient library that enables legacy socket applications to run without modification. Our experiments show that P-Socket achieves an end-to-end latency of 1.2μs and effective bandwidth of up to 2.87GB/s (out of a theoretical peak of 3.05GB/s).
- 1 November 2009
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 17-22
Variation-Aware Scheduling for Chip Multiprocessors with Thread Level Redundancy
Thread-level redundancy in Chip Multiprocessors(TLR-CMP) is efficient for soft error tolerance. Process variation causes core-to-core (C2C) performance asymmetry across a chip, which should be taken into consideration for application scheduling. In this paper, two types of variations beyond C2C are introduced, i.e., inter-pair and intra-pair variation in TLR-CMP. Intra-pair performance asymmetry can affect the performance of applications differently. Based on the above observation, we firstly formalize the variation aware scheduling in TLR-CMP as a 0-1 programming problem,to maximize the system weighted throughput. An efficient scheduling algorithm, named IntraVarF&AppSen, is then proposed to tackle this problem, which can be proved to be optimal when the number of applications to be scheduled is equal to the number of core pairs. Simulation on a 64-core CMP shows 2.8%-4% improvement in weighted throughput when compared to prior VarF&AppIPC algorithm.
- 1 November 2009
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 17-22
Variation-Aware Scheduling for Chip Multiprocessors with Thread Level Redundancy
Thread-level redundancy in Chip Multiprocessors(TLR-CMP) is efficient for soft error tolerance. Process variation causes core-to-core (C2C) performance asymmetry across a chip, which should be taken into consideration for application scheduling. In this paper, two types of variations beyond C2C are introduced, i.e., inter-pair and intra-pair variation in TLR-CMP. Intra-pair performance asymmetry can affect the performance of applications differently. Based on the above observation, we firstly formalize the variation aware scheduling in TLR-CMP as a 0-1 programming problem,to maximize the system weighted throughput. An efficient scheduling algorithm, named IntraVarF&AppSen, is then proposed to tackle this problem, which can be proved to be optimal when the number of applications to be scheduled is equal to the number of core pairs. Simulation on a 64-core CMP shows 2.8%-4% improvement in weighted throughput when compared to prior VarF&AppIPC algorithm.