Recently, the Architecture and Embedded Research Center of the School of Computer Science and Technology at Shandong University has achieved notable breakthroughs in big data storage. Three of their papers have been accepted at leading academic conferences. Specifically, the papers "HiDPU: A DPU-Oriented Hybrid Indexing Scheme for Disaggregated Storage Systems" and "HaSiS: A Hardware-assisted Single-index Store for Hybrid Transactional and Analytical Processing" have been accepted by FAST 2025 (USENIX Conference on File and Storage Technologies), a premier conference in the storage systems domain. Additionally, the paper "A Deep Dive into Protocol Design: How to Improve IPFS Performance without Sacrificing Decentralization" has been accepted by ICDE 2025 (IEEE International Conference on Data Engineering), one of the foremost conferences in the database field.

The research paper "HiDPU: A DPU-Oriented Hybrid Indexing Scheme for Disaggregated Storage Systems" is authored by Zhu Wenbin, a Ph.D. student at the School of Computer Science and Technology, Shandong University, with Professor Shen Zhaoyan as corresponding author and Professor Yu Dongxiao as co-author. HiDPU introduces a DPU (Data Processing Unit)-optimized hybrid indexing scheme for distributed storage systems. By utilizing a multi-level indexing structure, the scheme effectively addresses challenges such as limited DPU memory resources, constrained computational power, and high DPU-host interaction overhead. Additionally, the scheme incorporates an asynchronous index update strategy and compact upper-layer indexes, reducing interactions between the host and DPU while enhancing memory efficiency and query performance. The HiDPU prototype has been successfully implemented and validated on Huawei’s DPU to support petascale data addressing in disaggregated systems.

Fig 1 The hybrid indexing structure of HiDPU

Professor Shen Zhaoyan co-authored the research paper "HaSiS: A Hardware-assisted Single-index Store for Hybrid Transactional and Analytical Processing." HaSiS introduces an innovative single-index design that leverages computational storage drivers (CSDs) with transparent compression to effectively address the limitations of traditional multi-index approaches. This design ensures the freshness of OLAP (On-Line Analytical Processing) data and enhances the overall performance of both OLTP (On-Line Transaction Processing) and OLAP systems by utilizing sparsely populated storage blocks and an HTAP-oriented B+ tree structure.

The research paper "A Deep Dive into Protocol Design: How to Improve IPFS Performance Without Sacrificing Decentralization" is authored by Zhu Wenbin and the corresponding author is Professor Shen Zhaoyan, with Professors Yu Dongxiao and Zhao Mengying as co-authors. This study comprehensively analyzes the performance bottlenecks in the traditional IPFS (InterPlanetary File System) decentralized file protocol and introduces the optimized XIPFS protocol. The XIPFS protocol incorporates enhancements such as support for multi-node parallel block sharing, refined content publishing strategies, and improved node selection algorithms. These optimizations are fully compatible with decentralized protocols, allowing for enhanced node performance without compromising decentralization.

Fig 2 Hierarchical composition and operation flows of IPFS

FAST (Conference on File and Storage Technologies) was founded in 2002 and is co-organized by the USENIX Association and the ACM SIGOPS (Special Interest Group on Operating Systems). It is a leading international conference focused on storage technologies and is recognized as an A-level conference by the China Computer Federation (CCF), representing the research forefront in the field of computer storage. FAST typically attracts cutting-edge research contributions that address challenges in performance, scalability, reliability, and usability of storage systems, making it a key event for those interested in the future of file and storage technologies.

ICDE (IEEE International Conference on Data Engineering) is one of the premier conferences in the fields of databases and data engineering. It is also an A-level conference recommended by the China Computer Federation (CCF) and is ranked alongside SIGMOD and VLDB as one of the three leading conferences in database research. Organized by the IEEE Computer Society's Technical Committee on Data Engineering (TCDE), ICDE is a leading venue for presenting research and innovations in the design, implementation, and management of data systems and infrastructures.