Shun Gai
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Shun Gai has a total of 18 co-authors. Here are the top co-authors by number of shared publications:
- Yiming Zhang
Shanghai Jiao Tong University, Shanghai, China
3 shared publications - Jiwu Shu
Tsinghua University, Beijing, China
2 shared publications - Dongsheng Li
PDL Lab, College of Computer Science and Technology, National University of Defense Technology, Changsha, Hunan, China
1 shared publication - Guangtao Xue
Shanghai Jiao Tong University, Shanghai, China
1 shared publication - Haonan Wang
NICE XLab, Xiamen, Fujian, China
1 shared publication - Huiba Li
Alibaba Cloud, Beijing, China
1 shared publication
Latest Publications
- Open Access
- 28 January 2026
- Published by Association for Computing Machinery (ACM)
- p. 342-354
zBuffer: Zero-Copy and Metadata-Free Serialization for Fast RPC with Scatter-Gather Reflection
- 28 January 2026
- Published by Association for Computing Machinery (ACM)
- p. 342-354
Open AccesszBuffer: Zero-Copy and Metadata-Free Serialization for Fast RPC with Scatter-Gather Reflection
- Open Access
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 1126-1141
Cheetah: Metadata Aggregation for Fast Object Storage without Distributed Ordering
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 1126-1141
Open AccessCheetah: Metadata Aggregation for Fast Object Storage without Distributed Ordering
- 1 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 10, 2220-2225
SRockDB: A Range-Query Optimized Database Based on RocksDB
Data stores based on Log-Structured Merge Tree (LSM-Tree) are widely used in data centres, Artificial Learning and Machine Learning. As the core data structure, LSM-Tree is efficient in writing operations due to the out-of-place update and leveled design. However, it has limitations in writing, reading and space amplifications which decrease performance, especially range-query. Range query is one of the most important operations in data processing. To address this problem, we propose SRockDB. SRockDB is specifically designed to improve the performance of range queries by the in-memory cache. Compared to the block cache in RocksDB, SRockDB is more effective and coalesces adjacent keys to cache more data in limited memory. In experiments, SRockDB achieves 2.76× Queries Per Second (QPS), up to 41.74% lower tail latency, without downsides on the performance of write and point read.
- 1 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 10, 2220-2225
SRockDB: A Range-Query Optimized Database Based on RocksDB
Data stores based on Log-Structured Merge Tree (LSM-Tree) are widely used in data centres, Artificial Learning and Machine Learning. As the core data structure, LSM-Tree is efficient in writing operations due to the out-of-place update and leveled design. However, it has limitations in writing, reading and space amplifications which decrease performance, especially range-query. Range query is one of the most important operations in data processing. To address this problem, we propose SRockDB. SRockDB is specifically designed to improve the performance of range queries by the in-memory cache. Compared to the block cache in RocksDB, SRockDB is more effective and coalesces adjacent keys to cache more data in limited memory. In experiments, SRockDB achieves 2.76× Queries Per Second (QPS), up to 41.74% lower tail latency, without downsides on the performance of write and point read.
- Open Access
- 19 January 2023
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Storage
- Vol. 19 (1) , 1-22
Oasis : Controlling Data Migration in Expansion of Object-based Storage Systems
Object-based storage systems have been widely used for various scenarios such as file storage, block storage, blob (e.g., large videos) storage, and so on, where the data is placed among a large number of object storage devices (OSDs). Data placement is critical for the scalability of decentralized object-based storage systems. The state-of-the-art CRUSH placement method is a decentralized algorithm that deterministically places object replicas onto storage devices without relying on a central directory. While enjoying the benefits of decentralization such as high scalability, robustness, and performance, CRUSH-based storage systems suffer from uncontrolled data migration when expanding the capacity of the storage clusters (i.e., adding new OSDs), which is determined by the nature of CRUSH and will cause significant performance degradation when the expansion is nontrivial. This article presents MapX , a novel extension to CRUSH that uses an extra time-dimension mapping (from object creation times to cluster expansion times) for controlling data migration after cluster expansions. Each expansion is viewed as a new layer of the CRUSH map represented by a virtual node beneath the CRUSH root. MapX controls the mapping from objects onto layers by manipulating the timestamps of the intermediate placement groups (PGs). MapX is applicable to a large variety of object-based storage scenarios where object timestamps can be maintained as higher-level metadata. We have applied MapX to the state-of-the-art Ceph-RBD (RADOS Block Device) to implement a migration-controllable, decentralized object-based block store (called Oasis ). Oasis extends the RBD metadata structure to maintain and retrieve approximate object creation times (for migration control) at the granularity of expansion layers. Experimental results show that the MapX -based Oasis block store outperforms the CRUSH-based Ceph-RBD (which is busy in migrating objects after expansions) by 3.17× ∼ 4.31× in tail latency, and 76.3% (respectively, 83.8%) in IOPS for reads (respectively, writes).
- 19 January 2023
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Storage
- Vol. 19 (1) , 1-22
Open AccessOasis : Controlling Data Migration in Expansion of Object-based Storage Systems
Object-based storage systems have been widely used for various scenarios such as file storage, block storage, blob (e.g., large videos) storage, and so on, where the data is placed among a large number of object storage devices (OSDs). Data placement is critical for the scalability of decentralized object-based storage systems. The state-of-the-art CRUSH placement method is a decentralized algorithm that deterministically places object replicas onto storage devices without relying on a central directory. While enjoying the benefits of decentralization such as high scalability, robustness, and performance, CRUSH-based storage systems suffer from uncontrolled data migration when expanding the capacity of the storage clusters (i.e., adding new OSDs), which is determined by the nature of CRUSH and will cause significant performance degradation when the expansion is nontrivial. This article presents MapX , a novel extension to CRUSH that uses an extra time-dimension mapping (from object creation times to cluster expansion times) for controlling data migration after cluster expansions. Each expansion is viewed as a new layer of the CRUSH map represented by a virtual node beneath the CRUSH root. MapX controls the mapping from objects onto layers by manipulating the timestamps of the intermediate placement groups (PGs). MapX is applicable to a large variety of object-based storage scenarios where object timestamps can be maintained as higher-level metadata. We have applied MapX to the state-of-the-art Ceph-RBD (RADOS Block Device) to implement a migration-controllable, decentralized object-based block store (called Oasis ). Oasis extends the RBD metadata structure to maintain and retrieve approximate object creation times (for migration control) at the granularity of expansion layers. Experimental results show that the MapX -based Oasis block store outperforms the CRUSH-based Ceph-RBD (which is busy in migrating objects after expansions) by 3.17× ∼ 4.31× in tail latency, and 76.3% (respectively, 83.8%) in IOPS for reads (respectively, writes).