Shuai Wang
Shuai Wang has a total of 46 co-authors. Here are the top co-authors by number of shared publications:
- Dan Li
Tsinghua University, Beijing, Beijing, China
8 shared publications - Li Chen
Zhongguancun Laboratory, Beijing, Beijing, China
5 shared publications - Kaihui Gao
Zhongguancun Laboratory, Beijing, China
3 shared publications - Ennan Zhai
Alibaba Group, Hangzhou, China
2 shared publications - Zhentian Huang
Tsinghua University, Beijing, China
2 shared publications Binzhang Fu
Alibaba Group, Hangzhou, China
1 shared publication
Latest Publications
- Open Access
- 11 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 787-808
Networked Agent Memory and Causality Representation: Experiences towards Interpretable Cloud-Scale Root-Causing
- 11 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 787-808
Open AccessNetworked Agent Memory and Causality Representation: Experiences towards Interpretable Cloud-Scale Root-Causing
- Open Access
- 5 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 73-81
OptiFlow: Towards LLM-Driven Optimization of Collective Communication Algorithms
- 5 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 73-81
Open AccessOptiFlow: Towards LLM-Driven Optimization of Collective Communication Algorithms
- Open Access
- 30 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-2
Preventing Time-Shifting Attacks by Powerful Attackers with Time Public Key Infrastructure
- 30 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-2
Open AccessPreventing Time-Shifting Attacks by Powerful Attackers with Time Public Key Infrastructure
- Open Access
- 30 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 59-60
ACM CoNEXT 2025 Student Workshop (CoNEXT–SW'25): Workshop Summary
- 30 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 59-60
Open AccessACM CoNEXT 2025 Student Workshop (CoNEXT–SW'25): Workshop Summary
- Open Access
- 28 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 493-510
Measuring the Time Source Vulnerabilities in the NTP Ecosystem
- 28 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 493-510
Open AccessMeasuring the Time Source Vulnerabilities in the NTP Ecosystem
- Open Access
- 27 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 1269-1271
Discovering Millions of New Nodes and Links in the Internet by Challenging the Uniformity Assumption in Multipath Detection
- 27 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 1269-1271
Open AccessDiscovering Millions of New Nodes and Links in the Internet by Challenging the Uniformity Assumption in Multipath Detection
- Open Access
- 10 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 1135-1137
Poster: Q-Scanner: A Fast Scanning Tool for Large-Scale SSL/TLS Configurations Measurement
- 10 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 1135-1137
Open AccessPoster: Q-Scanner: A Fast Scanning Tool for Large-Scale SSL/TLS Configurations Measurement
- Open Access
- 29 June 2023
- Published by Association for Computing Machinery (ACM)
- p. 21-27
sRDMA: A General and Low-Overhead Scheduler for RDMA
- 29 June 2023
- Published by Association for Computing Machinery (ACM)
- p. 21-27
Open AccesssRDMA: A General and Low-Overhead Scheduler for RDMA
- 13 January 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ACM Transactions on Networking
- Vol. 31 (4) , 1748-1764
Dependable Virtualized Fabric on Programmable Data Plane
In modern multi-tenant data centers, each tenant desires reassuring dependability from the virtualized network fabric – bandwidth guarantee with work conservation, bounded tail latency and resilient reachability. However, the slow convergence of prior works under network dynamics and uncertainties can hardly provide the dependability for tenants. Further, state-of-the-art load balance schemes are guarantee-agnostic and bring great risks on breaking bandwidth guarantee, which is overlooked in prior works. In this paper, we propose, a dependable virtualized fabric framework which can (1) quickly detect network failure in data plane, (2) explicitly select proper paths for all flows, and (3) converge to ideal bandwidth allocation at sub-millisecond. The core idea of is to leverage the programmable data plane to build a fusion of an active edge (e.g., NIC) and an informative core (e.g., switch), where the core sends link status and tenant information to the edge via telemetry to help the latter make a timely and accurate decision on path selection and traffic admission. We fully implement with commodity SmartNICs and programmable switches. Extensive evaluations show that can keep bandwidth guarantee with high bandwidth utilization, low and bounded latency, and resilient reachability under various network scenarios with limited overhead. Application-level experiments show that can improve QPS by 2.4 $\times$ and cut tail latency by 10 $\times$ compared to the alternatives.
- 13 January 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ACM Transactions on Networking
- Vol. 31 (4) , 1748-1764
Dependable Virtualized Fabric on Programmable Data Plane
In modern multi-tenant data centers, each tenant desires reassuring dependability from the virtualized network fabric – bandwidth guarantee with work conservation, bounded tail latency and resilient reachability. However, the slow convergence of prior works under network dynamics and uncertainties can hardly provide the dependability for tenants. Further, state-of-the-art load balance schemes are guarantee-agnostic and bring great risks on breaking bandwidth guarantee, which is overlooked in prior works. In this paper, we propose, a dependable virtualized fabric framework which can (1) quickly detect network failure in data plane, (2) explicitly select proper paths for all flows, and (3) converge to ideal bandwidth allocation at sub-millisecond. The core idea of is to leverage the programmable data plane to build a fusion of an active edge (e.g., NIC) and an informative core (e.g., switch), where the core sends link status and tenant information to the edge via telemetry to help the latter make a timely and accurate decision on path selection and traffic admission. We fully implement with commodity SmartNICs and programmable switches. Extensive evaluations show that can keep bandwidth guarantee with high bandwidth utilization, low and bounded latency, and resilient reachability under various network scenarios with limited overhead. Application-level experiments show that can improve QPS by 2.4 $\times$ and cut tail latency by 10 $\times$ compared to the alternatives.
- 8 October 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ACM Transactions on Networking
- Vol. 30 (2) , 572-585
Impact of Synchronization Topology on DML Performance: Both Logical Topology and Physical Topology
To tackle the increasingly larger training data and models, researchers and engineers resort to multiple servers in a data center for distributed machine learning (DML). On one hand, DML enables us to leverage the computation power of multiple servers, which can effectively accelerate those computation-intensive tasks. On the other hand, DML also incurs significant communication cost due to parameter synchronization among these servers. In this paper, we want to explore the impact of synchronization topology, including both logical topology and physical topology, on the DML performance. First, we revisit the existing logical topologies, e.g., parameter server and ring allreduce, for parameter synchronization, and we find that these flat synchronization topologies is inefficient when running a large-scale DML training. Therefore, we propose a hierarchical parameter synchronization topology, called HiPS, which can achieve efficient parameter synchronization even on a large scale. Then, we compare two representative physical network topologies, namely, Fat-Tree and BCube. Based on our analyses, BCube has many advantages over Fat-Tree, e.g., higher bandwidth, better load balance, and lower hardware cost. The simulation results also show that BCube is more friendly to RDMA. Relying on the advantages of HiPS and BCube, the GST of “HiPS+BCube” is 12% ~ 70% lower than other combinations. Moreover, when the cluster size increases from 16 to 1024, the performance of “HiPS+BCube” only drops by 6.5%, while the performance of “Ring+BCube” drops by 44.6%. Hence, we believe “HiPS+BCube” is the optimal solution to benefit DML in large scale.
- 8 October 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ACM Transactions on Networking
- Vol. 30 (2) , 572-585
Impact of Synchronization Topology on DML Performance: Both Logical Topology and Physical Topology
To tackle the increasingly larger training data and models, researchers and engineers resort to multiple servers in a data center for distributed machine learning (DML). On one hand, DML enables us to leverage the computation power of multiple servers, which can effectively accelerate those computation-intensive tasks. On the other hand, DML also incurs significant communication cost due to parameter synchronization among these servers. In this paper, we want to explore the impact of synchronization topology, including both logical topology and physical topology, on the DML performance. First, we revisit the existing logical topologies, e.g., parameter server and ring allreduce, for parameter synchronization, and we find that these flat synchronization topologies is inefficient when running a large-scale DML training. Therefore, we propose a hierarchical parameter synchronization topology, called HiPS, which can achieve efficient parameter synchronization even on a large scale. Then, we compare two representative physical network topologies, namely, Fat-Tree and BCube. Based on our analyses, BCube has many advantages over Fat-Tree, e.g., higher bandwidth, better load balance, and lower hardware cost. The simulation results also show that BCube is more friendly to RDMA. Relying on the advantages of HiPS and BCube, the GST of “HiPS+BCube” is 12% ~ 70% lower than other combinations. Moreover, when the cluster size increases from 16 to 1024, the performance of “HiPS+BCube” only drops by 6.5%, while the performance of “Ring+BCube” drops by 44.6%. Hence, we believe “HiPS+BCube” is the optimal solution to benefit DML in large scale.