Yubo Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Yubo Liu has a total of 36 co-authors. Here are the top co-authors by number of shared publications:
- Zhiguang Chen
Sun Yat-sen University
5 shared publications - Yongfeng Wang
Sun Yat-sen University
3 shared publications - Yutong Lu
Sun Yat-sen University
3 shared publications - Hanjun Guo
Huawei Technologies, China
2 shared publications - Mingrui Liu
Huawei Technologies, China
2 shared publications Nong Xiao
School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou, China
2 shared publications
Latest Publications
- Open Access
- 18 November 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 24 (1) , 1-27
StrataJ: Enhancing File System Journaling Scalability via Adaptive Concurrency Control and Full-Stack Optimizations
The scalability of file system journaling faces significant challenges due to the increasing concurrency demands of modern multi-core CPUs and high throughput NVMe SSDs. However, our study reveals that the concurrency of journaling should be carefully adjusted to the workload, and the underlying I/O stack must be customized to better support concurrent journaling. To address these challenges, we propose StrataJ , a full-stack journaling system designed to enhance scalability through cross-layered optimizations spanning the journaling, block I/O, and driver layers. First, we design an adaptive journaling layer that dynamically adjusts the number of journaling streams based on the real-time analysis of workloads. Second, we propose a scalable I/O control mechanism in the block layer, which eliminates contention by parallelizing writeback throttling and physical region page (PRP) management. Third, we develop a decentralized I/O virtual address (IOVA) management scheme in the driver layer, which enables different regions of the address space to flush cache independently, thus in parallel. Our evaluations show that StrataJ significantly outperforms state-of-the-art systems across varied workloads, achieving up to 3.12×, 1.59× and 2.30× speedups on FxMark, Filebench, and RocksDB, respectively.
- 18 November 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 24 (1) , 1-27
Open AccessStrataJ: Enhancing File System Journaling Scalability via Adaptive Concurrency Control and Full-Stack Optimizations
The scalability of file system journaling faces significant challenges due to the increasing concurrency demands of modern multi-core CPUs and high throughput NVMe SSDs. However, our study reveals that the concurrency of journaling should be carefully adjusted to the workload, and the underlying I/O stack must be customized to better support concurrent journaling. To address these challenges, we propose StrataJ , a full-stack journaling system designed to enhance scalability through cross-layered optimizations spanning the journaling, block I/O, and driver layers. First, we design an adaptive journaling layer that dynamically adjusts the number of journaling streams based on the real-time analysis of workloads. Second, we propose a scalable I/O control mechanism in the block layer, which eliminates contention by parallelizing writeback throttling and physical region page (PRP) management. Third, we develop a decentralized I/O virtual address (IOVA) management scheme in the driver layer, which enables different regions of the address space to flush cache independently, thus in parallel. Our evaluations show that StrataJ significantly outperforms state-of-the-art systems across varied workloads, achieving up to 3.12×, 1.59× and 2.30× speedups on FxMark, Filebench, and RocksDB, respectively.
- Open Access
- 10 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 77-85
Towards Rack-as-a-Computer in Memory Interconnect Era with Coordinated Operating System Sharing
- 10 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 77-85
Open AccessTowards Rack-as-a-Computer in Memory Interconnect Era with Coordinated Operating System Sharing
- Open Access
- 11 February 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Storage
- Vol. 21 (2) , 1-26
WALSH: Write-Aggregating Log-Structured Hashing for Hybrid Memory
Persistent memory (PM) brings important opportunities for improving data storage including the widely used hash tables. However, PM is not friendly to small writes, which causes existing PM hashes to suffer from high hardware write amplification. Hybrid memory offers the performance and concurrency of DRAM and the durability and capacity of PM, but existing hybrid memory hashes cannot deliver high performance, low DRAM footprint, and fast recovery at the same time. This paper proposes WALSH, a flat hash with novel log-structured separate chaining designs to optimize the performance while ensuring low DRAM footprint and fast recovery. To address the overhead of hash resizing and garbage collection (GC), WALSH further proposes partial resizing/GC mechanisms and a 4-phase protocol for concurrent hash operations. As a result, WALSH is the first flat index for hybrid memory with embedded write aggregation ability. A comprehensive evaluation shows that WALSH substantially outperforms state-of-the-art hybrid memory hashes; e.g., its insert throughput is up to 2.4X that of related works while saving more than 87% of DRAM. WALSH also provides efficient recovery; e.g., it can recover a dataset with 1 billion objects in just a few seconds.
- 11 February 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Storage
- Vol. 21 (2) , 1-26
Open AccessWALSH: Write-Aggregating Log-Structured Hashing for Hybrid Memory
Persistent memory (PM) brings important opportunities for improving data storage including the widely used hash tables. However, PM is not friendly to small writes, which causes existing PM hashes to suffer from high hardware write amplification. Hybrid memory offers the performance and concurrency of DRAM and the durability and capacity of PM, but existing hybrid memory hashes cannot deliver high performance, low DRAM footprint, and fast recovery at the same time. This paper proposes WALSH, a flat hash with novel log-structured separate chaining designs to optimize the performance while ensuring low DRAM footprint and fast recovery. To address the overhead of hash resizing and garbage collection (GC), WALSH further proposes partial resizing/GC mechanisms and a 4-phase protocol for concurrent hash operations. As a result, WALSH is the first flat index for hybrid memory with embedded write aggregation ability. A comprehensive evaluation shows that WALSH substantially outperforms state-of-the-art hybrid memory hashes; e.g., its insert throughput is up to 2.4X that of related works while saving more than 87% of DRAM. WALSH also provides efficient recovery; e.g., it can recover a dataset with 1 billion objects in just a few seconds.
- 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
- 23 May 2022
- Published by Springer Nature in CCF Transactions on High Performance Computing
- Vol. 4 (3) , 321-338
A tail-tolerant cloud storage scheduling based on precise periodicity detection
Cloud storage is a fundamental component of the cloud computing system, which significantly affects the overall performance and quality of service of the cloud. Cloud storage servers face the challenge of imbalanced workloads. According to our observations on the time series generated by cloud storage, we found that the imbalance workloads will dramatically increase the tail latency of data access in the multi-tenant scenario. The intuitive solution is to periodicity detect the imbalance storage nodes and re-balance the loads. However, there are four challenges to accurately detect load of storage in the cloud with multiple tenants since the load may change frequently in cloud. This paper proposes PrecisePeriod, a precise periodicity detection algorithm customized for multi-tenant cloud storage. It removes outliers through data preprocessing, employs the discrete wavelet transform to remove high-frequency noise while keeping frequency domain information, computes the candidate periodicity queue using the autocorrelation function, and determines precise period through periodicity verification. Then, we design a cloud storage load balancing scheduling strategy based on PrecisePeriod, and the evaluation shows that the PrecisePeriod scheduling significantly reduces tail latency while only bringing $$1-2\%$$ overhead.- 23 May 2022
- Published by Springer Nature in CCF Transactions on High Performance Computing
- Vol. 4 (3) , 321-338
Open AccessA tail-tolerant cloud storage scheduling based on precise periodicity detection
Cloud storage is a fundamental component of the cloud computing system, which significantly affects the overall performance and quality of service of the cloud. Cloud storage servers face the challenge of imbalanced workloads. According to our observations on the time series generated by cloud storage, we found that the imbalance workloads will dramatically increase the tail latency of data access in the multi-tenant scenario. The intuitive solution is to periodicity detect the imbalance storage nodes and re-balance the loads. However, there are four challenges to accurately detect load of storage in the cloud with multiple tenants since the load may change frequently in cloud. This paper proposes PrecisePeriod, a precise periodicity detection algorithm customized for multi-tenant cloud storage. It removes outliers through data preprocessing, employs the discrete wavelet transform to remove high-frequency noise while keeping frequency domain information, computes the candidate periodicity queue using the autocorrelation function, and determines precise period through periodicity verification. Then, we design a cloud storage load balancing scheduling strategy based on PrecisePeriod, and the evaluation shows that the PrecisePeriod scheduling significantly reduces tail latency while only bringing $$1-2\%$$ overhead. - 14 October 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 614-619
GRSiamFC: Group Residual Convolutional Siamese Networks for Object Tracking
Visual tracking has seen considerable success with fully convolutional Siamese networks (SiamFC). However, the features extracted by SiamFC are redundant in practical applications. Most of the feature channels are ineffective or even serve as noise, which adversely affects tracking performance. The convolutional kernels tend to converge during the training phase of SiamFC, thereby the output features are typically redundant or irrelevant. In this paper, we propose a group residual convolutional Siamese network (GRSiamFC) to address this problem. Specifically, GRSiamFC divides the convolutional layers used for feature extraction into groups and trains each group in turn. Then the residuals between the output of previous groups and the true labels are used as losses to guide the current group to learn the residuals that were not fitted by the previous group, thereby improving the discriminative ability of the model. Extensive experiments on OTB50, OTB100, UAV123, and UAV20L show that the proposed GRSiamFC significantly improves the performance of SiamFC while keeping the complexity unchanged.
- 14 October 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 614-619
GRSiamFC: Group Residual Convolutional Siamese Networks for Object Tracking
Visual tracking has seen considerable success with fully convolutional Siamese networks (SiamFC). However, the features extracted by SiamFC are redundant in practical applications. Most of the feature channels are ineffective or even serve as noise, which adversely affects tracking performance. The convolutional kernels tend to converge during the training phase of SiamFC, thereby the output features are typically redundant or irrelevant. In this paper, we propose a group residual convolutional Siamese network (GRSiamFC) to address this problem. Specifically, GRSiamFC divides the convolutional layers used for feature extraction into groups and trains each group in turn. Then the residuals between the output of previous groups and the true labels are used as losses to guide the current group to learn the residuals that were not fitted by the previous group, thereby improving the discriminative ability of the model. Extensive experiments on OTB50, OTB100, UAV123, and UAV20L show that the proposed GRSiamFC significantly improves the performance of SiamFC while keeping the complexity unchanged.
- 1 September 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 167-176
KBP: Mining Block Access Pattern for I/O Prediction with K-Truss
Block prefetching is a technology widely used to improve the I/O efficiency of storage systems. Block access pattern prediction is a key part of the prefetching algorithm. However, existing block access pattern prediction methods cannot achieve the goals of low overhead, real-time, and self- adaptability at the same time. In this paper, we propose a real-time prediction method, called KBP (K-Truss-based Block access pattern Prediction). KBP uses SA (Sequential Access) Filter to identify and filter sequential access patterns to distinguish different patterns. Then, KBP uses K-Truss dense subgraph algorithm to detect compound access patterns, so as to use the time and space advantages of the K-Truss algorithm to reduce the overhead of access pattern recognition and make it possible for KBP to run in real time. Furthermore, KBP uses online reinforcement learning to achieve the goal of self- adaptability. We evaluate KBP in real-world workloads and the results show that KBP can improve the hit rate of 11.2% compared to the state-of-the-art prefetching algorithms on average.
- 1 September 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 167-176
KBP: Mining Block Access Pattern for I/O Prediction with K-Truss
Block prefetching is a technology widely used to improve the I/O efficiency of storage systems. Block access pattern prediction is a key part of the prefetching algorithm. However, existing block access pattern prediction methods cannot achieve the goals of low overhead, real-time, and self- adaptability at the same time. In this paper, we propose a real-time prediction method, called KBP (K-Truss-based Block access pattern Prediction). KBP uses SA (Sequential Access) Filter to identify and filter sequential access patterns to distinguish different patterns. Then, KBP uses K-Truss dense subgraph algorithm to detect compound access patterns, so as to use the time and space advantages of the K-Truss algorithm to reduce the overhead of access pattern recognition and make it possible for KBP to run in real time. Furthermore, KBP uses online reinforcement learning to achieve the goal of self- adaptability. We evaluate KBP in real-world workloads and the results show that KBP can improve the hit rate of 11.2% compared to the state-of-the-art prefetching algorithms on average.
- Open Access
- 30 January 2021
- Published by Springer Nature in Journal of Computer Science and Technology
- Vol. 36 (1) , 44-55
A GPU-Accelerated In-Memory Metadata Management Scheme for Large-Scale Parallel File Systems
- 30 January 2021
- Published by Springer Nature in Journal of Computer Science and Technology
- Vol. 36 (1) , 44-55
Open AccessA GPU-Accelerated In-Memory Metadata Management Scheme for Large-Scale Parallel File Systems
- Open Access
- 31 March 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 61289-61298
JCache: Journaling-Aware Flash Caching
Virtual machines (VMs) have become the predominant paradigm for deploying applications, and flash-based caches are widely deployed for VMs to improve their performance. However, due to the semantics of journaling in the VMs’ file system are transparent to flash caches, traditional flash caching with journaling induces inefficient logging I/O traffic to the shared storage server and duplicated caching in flash, which not only underutilizes flash caches but also aggravates their wearout. This paper presents JCache, a journaling-aware flash caching solution to address these problems. First, JCache proposes the virtual journal device design to receive and deliver the journaling semantics in the VMs to the cache manager so as to optimize the VM’s journaling. Second, JCache proposes a cache-only logging mechanism, which transparently uses the persistent flash caches as the journal area and does cache-only logging to safely eliminate the logging I/O traffic to the shared storage server. Third, JCache proposes the logical caching mechanism, which eliminates the duplicated flash caching induced by logging I/Os and in-place updates to mitigate the wearout of flash caches. Evaluations show that JCache improves the application performance by up to 11.4× and reduces the flash cache writes by up to 42% compared to traditional flash caching solutions.
- 31 March 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 61289-61298
Open AccessJCache: Journaling-Aware Flash Caching
Virtual machines (VMs) have become the predominant paradigm for deploying applications, and flash-based caches are widely deployed for VMs to improve their performance. However, due to the semantics of journaling in the VMs’ file system are transparent to flash caches, traditional flash caching with journaling induces inefficient logging I/O traffic to the shared storage server and duplicated caching in flash, which not only underutilizes flash caches but also aggravates their wearout. This paper presents JCache, a journaling-aware flash caching solution to address these problems. First, JCache proposes the virtual journal device design to receive and deliver the journaling semantics in the VMs to the cache manager so as to optimize the VM’s journaling. Second, JCache proposes a cache-only logging mechanism, which transparently uses the persistent flash caches as the journal area and does cache-only logging to safely eliminate the logging I/O traffic to the shared storage server. Third, JCache proposes the logical caching mechanism, which eliminates the duplicated flash caching induced by logging I/Os and in-place updates to mitigate the wearout of flash caches. Evaluations show that JCache improves the application performance by up to 11.4× and reduces the flash cache writes by up to 42% compared to traditional flash caching solutions.
- 11 December 2019
- Published by Springer Nature in Cluster Computing
- Vol. 23 (4) , 2501-2515
HasFS: optimizing file system consistency mechanism on NVM-based hybrid storage architecture
- 11 December 2019
- Published by Springer Nature in Cluster Computing
- Vol. 23 (4) , 2501-2515
HasFS: optimizing file system consistency mechanism on NVM-based hybrid storage architecture