Heng Pan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Heng Pan has a total of 90 co-authors. Here are the top co-authors by number of shared publications:
- Zhenyu Li
Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China and University of Chinese Academy of Sciences, Beijing, China
13 shared publications - Gaogang Xie
Chinese Academy of Sciences
8 shared publications - Kai Lv
Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China
4 shared publications Gareth Tyson
Hong Kong University of Science and Technology (GZ), Hong Kong
3 shared publications- Luyang Li
Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China and University of Chinese Academy of Sciences, Beijing, China
3 shared publications - Penghao Zhang
Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China
3 shared publications
Latest Publications
- Open Access
- 14 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 2074-2088
TurboRetry: Mitigating Large-Scale QUIC Handshake Floods with Off-the-Shelf DPU Offloading
- 14 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 2074-2088
Open AccessTurboRetry: Mitigating Large-Scale QUIC Handshake Floods with Off-the-Shelf DPU Offloading
- Open Access
- 11 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 1095-1108
GeoOrchestra: Orchestrating Heterogeneous Geo-Distributed Training with Network-Aware Scheduling
- 11 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 1095-1108
Open AccessGeoOrchestra: Orchestrating Heterogeneous Geo-Distributed Training with Network-Aware Scheduling
- Open Access
- 5 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 122-129
Integrating AI Clusters into Virtual Private Cloud
While commodity NIC-based back-end AI networks offer ultra-high intra-cluster bandwidth for distributed training, their limited programmability and on-chip resources hinder the implementation of advanced VPC features such as fine-grained isolation and stateful security policies. Furthermore, access to resources within the VPC needs to be routed through the front-end DPU, which is shared by the scale-up domain. The mismatch between the front-end DPU’s bandwidth and the back-end requirements causes GPU underutilization when intensive VPC communication is required for content recommendation, AIGC, and federated learning workloads. We propose an architecture that decouples complex policy enforcement from high-speed packet forwarding to support VPC semantics on back-end NICs and enable front-end/back-end integration. Evaluations show near-full GPU utilization in our analytical model and 71 μ s P999 extra latency of the first packet, suggesting that commodity hardware can support both high-throughput AI training and flexible VPC features.
- 5 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 122-129
Open AccessIntegrating AI Clusters into Virtual Private Cloud
While commodity NIC-based back-end AI networks offer ultra-high intra-cluster bandwidth for distributed training, their limited programmability and on-chip resources hinder the implementation of advanced VPC features such as fine-grained isolation and stateful security policies. Furthermore, access to resources within the VPC needs to be routed through the front-end DPU, which is shared by the scale-up domain. The mismatch between the front-end DPU’s bandwidth and the back-end requirements causes GPU underutilization when intensive VPC communication is required for content recommendation, AIGC, and federated learning workloads. We propose an architecture that decouples complex policy enforcement from high-speed packet forwarding to support VPC semantics on back-end NICs and enable front-end/back-end integration. Evaluations show near-full GPU utilization in our analytical model and 71 μ s P999 extra latency of the first packet, suggesting that commodity hardware can support both high-throughput AI training and flexible VPC features.
- Open Access
- 4 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 99-105
SwiftShift: Accelerating QUIC Migration for Ultra-Low-Latency Interactive Media
- 4 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 99-105
Open AccessSwiftShift: Accelerating QUIC Migration for Ultra-Low-Latency Interactive Media
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network and Service Management
- Vol. 23, 7751-7765
Themis: An Adjustable Congestion Control Framework for Improving Video QoE
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network and Service Management
- Vol. 23, 7751-7765
Themis: An Adjustable Congestion Control Framework for Improving Video QoE
- Open Access
- 24 November 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Networking
- Vol. 3 (CoNEXT4) , 1-25
Taurus: Towards A High-Performance and Generic Congestion Control Framework for Datacenter Networks
Congestion control (CC) is crucial for datacenter networks (DCNs), and CC frameworks are proposed to enable users to easily deploy new algorithms tailored to diverse scenarios. The framework is desired to be high-performance and generic: (i) allows CC to achieve high throughput and low latency. (ii) supports various algorithms and congestion scenarios. However, prior works either suffer from performance limitations or lack sufficient generality. CCP experiences throughput degradation under heavy traffic, while DOCA-PCC improves performance using hardware but lacks support for detecting and mitigating host congestion. In this paper, we present Taurus, a high-performance and generic CC framework through hardware-software co-design. To this end, Taurus partitions CC functions into distinct tasks and maps them onto suitable hardware/software components while mitigating excessive interaction overhead. Specifically, Taurus designs a collaborative signal collection mechanism to support diverse congestion feedback, a type-aware message report engine to reduce communication overhead, and software built-in handlers to facilitate deployments. We have implemented a fully functional Taurus on commodity servers with FPGA-based NICs. Experimental results show that Taurus supports various CC algorithms in achieving their near-native performance. Compared to CCP, Taurus improves throughput by 32.3%, reduces latency by 96.4%, and lowers CPU overhead by 158.7%. Compared to DOCA-PCC, Taurus improves throughput by 9.3% and reduces latency by 28.8%.
- 24 November 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Networking
- Vol. 3 (CoNEXT4) , 1-25
Open AccessTaurus: Towards A High-Performance and Generic Congestion Control Framework for Datacenter Networks
Congestion control (CC) is crucial for datacenter networks (DCNs), and CC frameworks are proposed to enable users to easily deploy new algorithms tailored to diverse scenarios. The framework is desired to be high-performance and generic: (i) allows CC to achieve high throughput and low latency. (ii) supports various algorithms and congestion scenarios. However, prior works either suffer from performance limitations or lack sufficient generality. CCP experiences throughput degradation under heavy traffic, while DOCA-PCC improves performance using hardware but lacks support for detecting and mitigating host congestion. In this paper, we present Taurus, a high-performance and generic CC framework through hardware-software co-design. To this end, Taurus partitions CC functions into distinct tasks and maps them onto suitable hardware/software components while mitigating excessive interaction overhead. Specifically, Taurus designs a collaborative signal collection mechanism to support diverse congestion feedback, a type-aware message report engine to reduce communication overhead, and software built-in handlers to facilitate deployments. We have implemented a fully functional Taurus on commodity servers with FPGA-based NICs. Experimental results show that Taurus supports various CC algorithms in achieving their near-native performance. Compared to CCP, Taurus improves throughput by 32.3%, reduces latency by 96.4%, and lowers CPU overhead by 158.7%. Compared to DOCA-PCC, Taurus improves throughput by 9.3% and reduces latency by 28.8%.
- Open Access
- 6 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 135-141
OmniDMA: Scalable RDMA Transport over WAN
- 6 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 135-141
Open AccessOmniDMA: Scalable RDMA Transport over WAN
- Open Access
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 498-514
Harmonia: A Unified Framework for Heterogeneous FPGA Acceleration in the Cloud
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 498-514
Open AccessHarmonia: A Unified Framework for Heterogeneous FPGA Acceleration in the Cloud
- Open Access
- 22 April 2024
- Published by Association for Computing Machinery (ACM)
- p. 82-98
Hoda: a High-performance Open vSwitch Dataplane with Multiple Specialized Data Paths
- 22 April 2024
- Published by Association for Computing Machinery (ACM)
- p. 82-98
Open AccessHoda: a High-performance Open vSwitch Dataplane with Multiple Specialized Data Paths
- 10 March 2023
- Published by Springer Nature
- p. 227-253
Towards Diagnosing Accurately the Performance Bottleneck of Software-Based Network Function Implementation
- 10 March 2023
- Published by Springer Nature
- p. 227-253
Towards Diagnosing Accurately the Performance Bottleneck of Software-Based Network Function Implementation