Xinwei Hu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Xinwei Hu has a total of 33 co-authors. Here are the top co-authors by number of shared publications:
- Yuxin Ren
Huawei Technologies, Beijing, China
7 shared publications - Ning Jia
Huawei Technologies, Beijing, China
6 shared publications - Ding Li
Key Laboratory of High-Confidence Software Technologies (MOE), School of Computer Science, Peking University, Beijing, China
4 shared publications - Yao Guo
Peking University
4 shared publications - Peng Jiang
Southeast University, Nanjing, China
2 shared publications - Ruizhe Huang
Peking University, Beijing, China
2 shared publications
Latest Publications
- Open Access
- 27 September 2026
- Published by Association for Computing Machinery (ACM)
- p. 693-703
Governing Global ASID/PCID Management for QoS Isolation in Shared-Kernel Mixed-Priority Systems
- 27 September 2026
- Published by Association for Computing Machinery (ACM)
- p. 693-703
Open AccessGoverning Global ASID/PCID Management for QoS Isolation in Shared-Kernel Mixed-Priority Systems
- Open Access
- 30 June 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Software Engineering
- Vol. 3 (FSE) , 4001-4024
Unleashing HPC Application Performance through Software Deployment: A Joint Model of Software Parallelism and Co-location
Software deployment is a critical software engineering practice, particularly for high performance computing (HPC) software.The deployment determines the software execution performance because the deployment maps a number of software components to multiple CPUs in a server. Inappropriate mapping decreases software parallelism and increases resource contention due to software co-location on the same CPUs. However, calculating the mapping to maximize the software performance is challenging, primarily due to the lack of a joint performance model that accounts for both software parallelism and co-location. Consequently, existing industry practice has to rely on experienced engineers to manually tune the mapping during deployment, resulting in substantial human resource waste of man-months and suboptimal software performance. This paper proposes a holistic approach to mapping multiple CPUs among multiple software components to achieve better applicability and performance. We develop a performance model for predicting performance impact of different CPU mapping configurations, along with a search algorithm to identify the best mapping scheme. Our performance model jointly considers software parallelism and co-location, breaks the performance estimation into regularized execution and interference coefficient to improve accuracy, and integrates expert knowledge to reduce the model complexity. Our search algorithm employs nested iterative packing algorithm to explore all possible mapping schemes, thereby uncovering the optimal solution. Evaluation on a multi module HPC application shows 17% better performance than its default CPU mapping Our solution has been deployed in a commercial HPC cluster with more than 50K CPU cores, delivering 26.5% performance improvement and saving many man-months effort spent on performance tuning.
- 30 June 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Software Engineering
- Vol. 3 (FSE) , 4001-4024
Open AccessUnleashing HPC Application Performance through Software Deployment: A Joint Model of Software Parallelism and Co-location
Software deployment is a critical software engineering practice, particularly for high performance computing (HPC) software.The deployment determines the software execution performance because the deployment maps a number of software components to multiple CPUs in a server. Inappropriate mapping decreases software parallelism and increases resource contention due to software co-location on the same CPUs. However, calculating the mapping to maximize the software performance is challenging, primarily due to the lack of a joint performance model that accounts for both software parallelism and co-location. Consequently, existing industry practice has to rely on experienced engineers to manually tune the mapping during deployment, resulting in substantial human resource waste of man-months and suboptimal software performance. This paper proposes a holistic approach to mapping multiple CPUs among multiple software components to achieve better applicability and performance. We develop a performance model for predicting performance impact of different CPU mapping configurations, along with a search algorithm to identify the best mapping scheme. Our performance model jointly considers software parallelism and co-location, breaks the performance estimation into regularized execution and interference coefficient to improve accuracy, and integrates expert knowledge to reduce the model complexity. Our search algorithm employs nested iterative packing algorithm to explore all possible mapping schemes, thereby uncovering the optimal solution. Evaluation on a multi module HPC application shows 17% better performance than its default CPU mapping Our solution has been deployed in a commercial HPC cluster with more than 50K CPU cores, delivering 26.5% performance improvement and saving many man-months effort spent on performance tuning.
- Open Access
- 26 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 2141-2156
Practical and Scalable RDMA Connection Sharing for HPC Workload
- 26 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 2141-2156
Open AccessPractical and Scalable RDMA Connection Sharing for HPC Workload
- Open Access
- 19 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 269-282
Cost-Efficient Cloud Infrastructure with Hugepage-aware Memory Deduplication
- 19 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 269-282
Open AccessCost-Efficient Cloud Infrastructure with Hugepage-aware Memory Deduplication
- Open Access
- 23 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 15-21
Towards Efficient Hugepage-aware Memory Deduplication
Memory deduplication techniques aim to reduce memory consumption by sharing redundant memory pages. Additionally, hugepages are commonly employed to optimize performance by minimizing TLB misses. However, the simultaneous use of memory deduplication and hugepages presents a technical challenge. Existing approaches either split hugepages to achieve efficient memory deduplication or overlook redundant portions within hugepages to maintain high performance. As a result, they are compelled to compromise either performance or memory utilization. This paper identifies the fundamental issue causing the current dilemma as the absence of tight cooperation between the host and guest. Existing approaches manage all actions within the host OS without engaging the guest OS. In contrast, we propose a new memory deduplication framework, which adopts a host-guest co-design approach. We integrate host memory deduplication, shared data hot-plugging, and guest memory migration to simultaneously attain superior performance, a high deduplication rate, and reduced memory consumption.
- 23 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 15-21
Open AccessTowards Efficient Hugepage-aware Memory Deduplication
Memory deduplication techniques aim to reduce memory consumption by sharing redundant memory pages. Additionally, hugepages are commonly employed to optimize performance by minimizing TLB misses. However, the simultaneous use of memory deduplication and hugepages presents a technical challenge. Existing approaches either split hugepages to achieve efficient memory deduplication or overlook redundant portions within hugepages to maintain high performance. As a result, they are compelled to compromise either performance or memory utilization. This paper identifies the fundamental issue causing the current dilemma as the absence of tight cooperation between the host and guest. Existing approaches manage all actions within the host OS without engaging the guest OS. In contrast, we propose a new memory deduplication framework, which adopts a host-guest co-design approach. We integrate host memory deduplication, shared data hot-plugging, and guest memory migration to simultaneously attain superior performance, a high deduplication rate, and reduced memory consumption.
- Open Access
- 23 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 38-44
Cache or Direct Access? Revitalizing Cache in Heterogeneous Memory File System
This paper revisits the value of cache in DRAM-PM heterogeneous memory file systems. The first contribution is a comprehensive analysis of the existing file systems on heterogeneous memory, including cache-based and DAX-based (direct access). We find that the DRAM cache still plays an important role in heterogeneous memory. The second contribution is a cache framework for heterogeneous memory, called FLAC. FLAC integrates the cache with the virtual memory management and proposes two technologies of zero-copy caching and parallel-optimized cache management, which deliver the benefits of fast application-storage data transfer and efficient DRAM-PM data synchronization/migration. We further implement a library file system upon FLAC. Microbenchmarks show that FLAC provides a performance increase of up to two orders of magnitude over existing file systems in file read/write. With a real-world application, FLAC achieves up to 77.4% and 89.3% better performance than NOVA and EXT4, respectively.
- 23 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 38-44
Open AccessCache or Direct Access? Revitalizing Cache in Heterogeneous Memory File System
This paper revisits the value of cache in DRAM-PM heterogeneous memory file systems. The first contribution is a comprehensive analysis of the existing file systems on heterogeneous memory, including cache-based and DAX-based (direct access). We find that the DRAM cache still plays an important role in heterogeneous memory. The second contribution is a cache framework for heterogeneous memory, called FLAC. FLAC integrates the cache with the virtual memory management and proposes two technologies of zero-copy caching and parallel-optimized cache management, which deliver the benefits of fast application-storage data transfer and efficient DRAM-PM data synchronization/migration. We further implement a library file system upon FLAC. Microbenchmarks show that FLAC provides a performance increase of up to two orders of magnitude over existing file systems in file read/write. With a real-world application, FLAC achieves up to 77.4% and 89.3% better performance than NOVA and EXT4, respectively.
- Open Access
- 24 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 16-23
Towards OS Heterogeneity Aware Cluster Management for HPC
- 24 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 16-23
Open AccessTowards OS Heterogeneity Aware Cluster Management for HPC