XCXin Chen

Xin Chen

*This is an algorithmically generated author record based on publicly available information.

Xin Chen has a total of 36 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    High-Throughput LLM inference on Heterogeneous Clusters

    by Yi Xiong, Jinqi Huang, Wenjie Huang, Xuebing Yu, Entong Li, Zhixiong Ning, Jinhua Zhou, Li Zeng, Xin Chen

    Nowadays, many companies possess various types of AI accelerators, forming heterogeneous clusters. Efficiently leveraging these clusters for high-throughput large language model (LLM) inference services can significantly reduce costs and expedite task processing. However, LLM inference on heterogeneous clusters presents two main challenges. Firstly, different deployment configurations can result in vastly different performance. The number of possible configurations is large, and evaluating the effectiveness of a specific setup is complex. Thus, finding an optimal configuration is not an easy task. Secondly, LLM inference instances within a heterogeneous cluster possess varying processing capacities, leading to different processing speeds for handling inference requests. Evaluating these capacities and designing a request scheduling algorithm that fully maximizes the potential of each instance is challenging. In this paper, we propose a high-throughput inference service system on heterogeneous clusters. First, the deployment configuration is optimized by modeling the resource amount and expected throughput and using the exhaustive search method. Second, a novel mechanism is proposed to schedule requests among instances, which fully considers the different processing capabilities of various instances. Extensive experiments show that the proposed scheduler improves throughput by 122.5% and 33.6% on two heterogeneous clusters, respectively.

  • Open Access

    WindGP: Efficient Graph Partitioning on Heterogenous Machines

    by Li Zeng, Haohan Huang, Binfan Zheng, Kang Yang, Shengcheng Shao, Jinhua Zhou, Jun Xie, Rongqian Zhao, Xin Chen

    Graph Partitioning is widely used in many real-world applications such as fraud detection and social network analysis, in order to enable the distributed graph computing on large graphs. However, existing works fail to balance the computation cost and communication cost on machines with different power (including computing capability, network bandwidth and memory size), as they only consider replication factor and neglect the difference of machines in realistic data centers. In this paper, we propose a general graph partitioning algorithm WindGP, which can support fast and high-quality edge partitioning on heterogeneous machines. WindGP designs novel preprocessing techniques to simplify the metric and balance the computation cost according to the characteristics of graphs and machines. Also, best-first search is proposed instead of BFS and DFS, in order to generate clusters with high cohesion. Furthermore, WindGP adaptively tunes the partition results by sophisticated local search methods. Extensive experiments show that WindGP outperforms all state-of-the-art partition methods by 1.35 - 27 times on both dense and sparse distributed graph algorithms, and has good scalability with graph size and machine number.

  • Open Access

    Coordinating Multiple Intelligent Reflecting Surfaces without Channel Information

    by Fan Xu, Jiawei Yao, Wenhai Lai, Kaiming Shen, Xin Li, Xin Chen, Zhi-Quan Luo

    Conventional beamforming methods for intelligent reflecting surfaces (IRSs) or reconfigurable intelligent surfaces (RISs) typically entail the full channel state information (CSI). However, the computational cost of channel acquisition soars exponentially with the number of IRSs. To bypass this difficulty, we propose a novel strategy called blind beamforming that coordinates multiple IRSs by means of statistics without knowing CSI. Blind beamforming only requires measuring the received signal power at the user terminal for a sequence of randomly generated phase shifts across all IRSs. The main idea is to extract the key statistical quantity for beamforming by exploring only a small portion of the whole solution space of phase shifts. We show that blind beamforming guarantees a signal-to-noise ratio (SNR) boost of Theta(N^{2L}) under certain conditions, where L is the number of IRSs and N is the number of reflecting elements per IRS. The proposed conditions for achieving the optimal SNR boost of Theta(N^{4}) in a double-IRS system are much easier to satisfy than the existing ones in the literature. Most importantly, the proposed conditions can be extended to a fully general L-IRS system. The above result significantly improves upon the state of the art in the area of multi-IRS-assisted communication. Moreover, blind beamforming is justified via field tests and simulations. In particular, as shown in our field tests at 2.6 GHz, our method yields up to 17 dB SNR boost; to the best of our knowledge, this is the first time that the use of multiple IRSs gets verified in the real world.

  • KBQA: Accelerate Fuzzy Path Query on Knowledge Graph

    by Li Zeng, Qiheng You, Jincheng Lu, Shizheng Liu, Weijian Sun, Rongqian Zhao, Xin Chen


  • SQLG+: Efficient k-hop Query Processing on RDBMS

    by Li Zeng, Jinhua Zhou, Shijun Qin, Haoran Cai, Rongqian Zhao, Xin Chen


  • Open Access

    Sim-to-Real Optimization of Complex Real World Mobile Network with Imperfect Information via Deep Reinforcement Learning from Self-play

    by Yongxi Tan, Jin Yang, Xin Chen, Qitao Song, Yunjun Chen, Zhangxiang Ye, Zhenqiang Su

    Mobile network that millions of people use every day is one of the most complex systems in the world. Optimization of mobile network to meet exploding customer demand and reduce capital/operation expenditures poses great challenges. Despite recent progress, application of deep reinforcement learning (DRL) to complex real world problem still remains unsolved, given data scarcity, partial observability, risk and complex rules/dynamics in real world, as well as the huge reality gap between simulation and real world. To bridge the reality gap, we introduce a Sim-to-Real framework to directly transfer learning from simulation to real world via graph convolutional neural network (CNN) - by abstracting partially observable mobile network into graph, then distilling domain-variant irregular graph into domain-invariant tensor in locally Euclidean space as input to CNN -, domain randomization and multi-task learning. We use a novel self-play mechanism to encourage competition among DRL agents for best record on multiple tasks via simulated annealing, just like athletes compete for world record in decathlon. We also propose a decentralized multi-agent, competitive and cooperative DRL method to coordinate the actions of multi-cells to maximize global reward and minimize negative impact to neighbor cells. Using 6 field trials on commercial mobile networks, we demonstrate for the first time that a DRL agent can successfully transfer learning from simulation to complex real world problem with imperfect information, complex rules/dynamics, huge state/action space, and multi-agent interactions, without any training in the real world.