Yicheng Liu
I am a second-year Ph.D. student at UCLA, co-advised by Sam Kumar and Harry Xu.
I was an undergraduate student at Shanghai Jiao Tong University (SJTU).
When I was an undergraduate student at SJTU, I was an intern at Institution of Parallel And Distributed System (IPADS), advised by Jinyu Gu.
In the 2024 Summer, I was an intern in University of Washington, System Group, advised by Baris Kaciksi and mentored by Yigong Hu.
In the 2023 Summer, I visited and took part in the research in University of Michigan, OrderLab, advised by Ryan Huang.
Research Interests
Make system secure and reliable: Resource contention is awkward.
Some abnormal workloads leading the system to severe contention depresses the reliability.
Other malicious users making use of the contention to launch DoS attack harms the security.
I work toward analyzing and mitigating the contention with some novel mechanisms.
Make secure and reliable system faster: Security is expensive.
The software solution costs efficiency.
The hardware solution costs money.
I seek a co-design approach to accelerate current solutions while making minimal changes to existing hardware.
AI 4 SYS: I have experience in AI research and am eager to combine this expertise with system research.
Publications
Mitigating Application Resource Overload with Targeted Task Cancellation
Yigong Hu, Zeyin Zhang, Yicheng Liu, Yile Gu, Shuangyu Lei, Baris Kasikci, Peng Huang
ACM Symposium on Operating Systems Principles (SOSP 25)
Modern software must deliver both high SLO attainment and low request drop rate. However, achieving this balance is challenging due to subtle and unpredictable internal resource contention among concurrently executing requests. Traditional overload control mechanisms, which rely on global signals such as queuing delays, fail to handle application resource overload effectively because they cannot accurately predict which requests will monopolize critical resources. In this paper, we propose Atropos, an overload control framework that proactively cancels requests causing severe resource contention rather than victim requests that are blocked by it. Atropos continuously monitors the resource usage of executing requests, identifies the requests contributing most significantly to resource overload, and selectively cancels them. We integrate Atropos into six large-scale applications and evaluate it against 16 real-world overload scenarios. Our results demonstrate that Atropos successfully maintains application performance targets, significantly outperforming existing state-of-the-art solutions and highlighting its robustness and general applicability.
Chumeng Liang, Zherui Huang, Yicheng Liu, Zhanyu Liu, Guanjie Zheng, Hanyuan Shi, Kan Wu, Yuhao du, Fuliang Li, Zhenhui Jeesie Li
ACM SIGKDD Conference on Knowledge Discovery and Data Mining (SIGKDD 23)
abstract: "Traffic simulation provides interactive data for the optimization of traffic policies. However, existing traffic simulators are limited by their lack of scalability and shortage in input data, which prevents them from generating interactive data from traffic simulation in the scenarios of real large-scale city road networks. In this paper, we present City Brain Lab, a toolkit for scalable traffic simulation. CBLab is consist of three components: CBEngine, CBData, and CBScenario. CBEngine is a highly efficient simulators supporting large scale traffic simulation. CBData includes a traffic dataset with road network data of 100 cities all around the world. We also develop a pipeline to conduct one-click transformation from raw road networks to input data of our traffic simulation. Combining CBEngine and CBData allows researchers to run scalable traffic simulation in the road network of real large-scale cities. Based on that, CBScenario implements an interactive environment and several baseline methods for two scenarios of traffic policies respectively, with which traffic policies adaptable for large-scale urban traffic can be trained and tuned. To the best of our knowledge, CBLab is the first infrastructure supporting traffic policy optimization on large-scale urban scenarios."
Zefang Yu, Yangcheng Li, Yicheng Liu, Ting Liu, Yuzhuo Fu
International Conference on Acoustics, Speech and Signal Processing (ICASSP 22)
Deep learning-based methods for human pose estimation require large volumes of training data to achieve superior performance. However, data acquisition in classroom environments raises privacy concerns, which will undoubtedly hinder the development of the latest deep learning techniques in education domain. Due to the absence of large, richly annotated classroom datasets, research into classroom observation has had to be done by manually collecting and annotating datasets. Unfortunately, the annotation of such data is time-consuming and challenging in over-crowded classrooms. To break through these limitations, we open source SynPose, a large, densely labeled synthetic dataset specifically designed for crowded human pose estimation in classroom and meeting scenarios. Moreover, we propose a novel CTGAN to bridge the domain gap. Comprehensive experiments on real-world classroom images show that our proposed dataset and method deliver important performance benefits compared to existing datasets, revealing the potential of SynPose for future studies.
