Renchao Xie
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Renchao Xie has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
Hongliang Lu
Unit 61905 of PLA, Shenyang 110005, China
1 shared publication- Jiang Liu
Beijing University of Posts and Telecommunications, Beijing, China
1 shared publication Juan Liu
State Key Laboratory of Networking and Switching Technology Beijing University of Post and Telecommunications 100876 Beijing China
1 shared publicationLi Feng
State key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing 100876, China
1 shared publicationQinqin Tang
State key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing 100876, China
1 shared publicationTao Huang
State key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing 100876, China
1 shared publication
Latest Publications
- Open Access
- 30 November 2025
- Published by Fundamental and Frontier Science Publication Group Limited in Journal of Intelligent Computing and Networking
- Vol. 1 (2) , 65-78
Optimizing Distributed Training in Computing Power Networks: A Heterogeneity-aware Approach
Distributed training over computing power network (CPN) suffers from high round-trip times (RTTs), volatile bandwidth, heterogeneous accelerators, and non-independent identically distributed (non‑IID) data, which collectively lead to imbalanced collaboration and slow convergence. To address these issues, we propose a heterogeneity-aware distributed training framework tailored for wide area network (WAN) settings. It adopts a hybrid hierarchical synchronization mechanism that redistributes communication load temporally and spatially, thus improving end-to-end training efficiency. In addition, we design a heuristic algorithm for optimizing the synchronization topology, guided by the computational capacities of training nodes and real-time network telemetry. We implement a distributed training system and validate the proposed framework through extensive simulations, demonstrating consistent performance gains across diverse heterogeneous settings.
- 30 November 2025
- Published by Fundamental and Frontier Science Publication Group Limited in Journal of Intelligent Computing and Networking
- Vol. 1 (2) , 65-78
Open AccessOptimizing Distributed Training in Computing Power Networks: A Heterogeneity-aware Approach
Distributed training over computing power network (CPN) suffers from high round-trip times (RTTs), volatile bandwidth, heterogeneous accelerators, and non-independent identically distributed (non‑IID) data, which collectively lead to imbalanced collaboration and slow convergence. To address these issues, we propose a heterogeneity-aware distributed training framework tailored for wide area network (WAN) settings. It adopts a hybrid hierarchical synchronization mechanism that redistributes communication load temporally and spatially, thus improving end-to-end training efficiency. In addition, we design a heuristic algorithm for optimizing the synchronization topology, guided by the computational capacities of training nodes and real-time network telemetry. We implement a distributed training system and validate the proposed framework through extensive simulations, demonstrating consistent performance gains across diverse heterogeneous settings.
- Open Access
- 1 January 2016
- Published by National Library of Serbia in Computer Science and Information Systems
- Vol. 13 (2) , 547-560
Reliability-based controller placement algorithm in software defined networking
Software Defined Networking (SDN) change the routing architecture from distribution to centralization. Then the centralized controller placement becomes an important issue since network delay and synchronization will be affected. In this paper, the problem of controller placement to optimize the network average reliability is studied. We first consider the optimization problem for controller placement under the assumption with the shortest path between controller and switches, and a clustering based optimal controller placement algorithm and sub-optimal algorithm base on greedy algorithm are proposed, respectively. Then, we extend the controller placement problem to more general case with multi-paths between controller and switches. In this case, to reduce the computation complexity, a Reliability Factor is defined to replace the network average reliability. After that, modified clustering based and greedy based algorithms are proposed to make the optimal and sub-optimal solution, respectively. Finally, simulation results demonstrate the performance of the proposed algorithms better than regular placement stratagem with the Internet2 OS3E and the Internet Topology Zoo.
- 1 January 2016
- Published by National Library of Serbia in Computer Science and Information Systems
- Vol. 13 (2) , 547-560
Open AccessReliability-based controller placement algorithm in software defined networking
Software Defined Networking (SDN) change the routing architecture from distribution to centralization. Then the centralized controller placement becomes an important issue since network delay and synchronization will be affected. In this paper, the problem of controller placement to optimize the network average reliability is studied. We first consider the optimization problem for controller placement under the assumption with the shortest path between controller and switches, and a clustering based optimal controller placement algorithm and sub-optimal algorithm base on greedy algorithm are proposed, respectively. Then, we extend the controller placement problem to more general case with multi-paths between controller and switches. In this case, to reduce the computation complexity, a Reliability Factor is defined to replace the network average reliability. After that, modified clustering based and greedy based algorithms are proposed to make the optimal and sub-optimal solution, respectively. Finally, simulation results demonstrate the performance of the proposed algorithms better than regular placement stratagem with the Internet2 OS3E and the Internet Topology Zoo.