Henry Hong Xu
Also known as Hong Xu
Henry Hong Xu is a computer scientist at the Chinese University of Hong Kong (CUHK) whose research covers machine learning systems and data center networks. He earned a B.Eng. at CUHK in 2007 and a Ph.D. at the University of Toronto in 2013, then worked at City University of Hong Kong from 2013 to 2020. His coauthored NAssim paper won the ACM SIGCOMM Best Paper Award in 2022, and he coauthored Lina, which speeds up mixture-of-experts models, at USENIX ATC in 2023 and Mycroft, a tracing system for LLM training, at SOSP in 2025. A paper from his NetX Lab on SONiC DASH SmartSwitch received the NSDI Community Award in 2026. He is an Associate Professor in the Department of Computer Science and Engineering at CUHK.
- Badges
Timeline
-
2007–2013
Degrees at CUHK and Toronto
He earns a B.Eng. from The Chinese University of Hong Kong (CUHK) in 2007, then an M.A.Sc. in 2009 and a Ph.D. in 2013 from the University of Toronto.
-
2013–2020
Years at City University of Hong Kong
From 2013 to 2020 he is with City University of Hong Kong. In 2014 he receives an Early Career Scheme Grant from the Hong Kong Research Grants Council, and in 2015 his work wins an IEEE ICNP best paper award.
-
2022
A SIGCOMM best paper
In 2022, he is a coauthor of NAssim, a software-defined network assimilation system that wins the ACM SIGCOMM Best Paper Award. NAssim parses device manuals into configuration models, validates them, and applies a deep-learning-based mapping algorithm that matches them to the model in the SDN controller.
-
2023
Lina for mixture-of-experts models
In 2023, he coauthors Lina at USENIX ATC with Jiamin Li, Yimin Jiang, Yibo Zhu and Cong Wang. Lina speeds up distributed mixture-of-experts training and inference, cutting training step time by up to 1.73x and 95th-percentile inference time by an average of 1.63x, and ByteDance adopts it for training BERT-based MoE models.
-
2023
An ICNP best paper award
In October 2023, “Enabling Load Balancing for Lossless Datacenters”, a paper he coauthors, receives the IEEE ICNP 2023 Best Paper Award.
-
2025
Mycroft for LLM training
He coauthors Mycroft with Minlan Yu and others, published at SOSP in October 2025. Mycroft is a tracing and root cause analysis system for collective communication in LLM training; deployed at ByteDance, it detects anomalies within 15 seconds in 90% of cases.
-
2026
NSDI Community Award
In 2026, the paper “Offloading Cloud Network Services at Production Scale with SONiC DASH SmartSwitch”, led by Shaofeng Wu, a final-year Ph.D. student in his NetX Lab, receives the NSDI Community Award as one of four papers selected. The system is deployed in Microsoft Azure and reaches 1.53 Tbps throughput and 256 million concurrent connections.
-
As of October 2026
Associate Professor at CUHK
He is an Associate Professor in the Department of Computer Science and Engineering at CUHK, with research on networking and systems, particularly machine learning systems and data center networks. In February 2026, he gives a keynote on AI infrastructure at the NINeS 2026 HKUST pod and at the Secure & Scalable Machine Learning Symposium 2026 in Thailand.
Career
- NowChinese University of Hong KongAssociate Professor
- 2013–2020City University of Hong Kong
Education
- 2013University of TorontoPh.D.
- 2009University of TorontoM.A.Sc.
- 2007Chinese University of Hong KongB.Eng.
Sources and credits
Photo: We have no published photo of Henry Hong Xu yet. Send us a free-licensed photo.
- Henry Hong Xu – CUHK CSE
- Hong XU - The Chinese University of Hong Kong
- Prof. Hong Xu received the Best Paper Award from ACM SIGCOMM 2022 – CUHK CSE
- acm sigcomm - Best Paper Award
- Accelerating Distributed MoE Training and Inference with Lina | USENIX
- [2210.17223] Accelerating Distributed MoE Training and Inference with Lina
- Publications – Hong Xu
- [2509.03018] Mycroft: Tracing Dependencies in Collective Communication Towards Reliable LLM Training
- Publications – Minlan Yu
- Ph.D. Student Shaofeng Wu from Prof. Hong Xu’s Lab Receives NSDI’26 Community Award – CUHK CSE
- Offloading Cloud Network Services at Production Scale with SONiC DASH SmartSwitch | USENIX
- Hong Xu, Henry
