Data-Intensive Computing &
AI Systems Laboratory

Kihwan Kim, Hyunsun Chung, Seonghoon Ahn, Junhyeok Park, Safdar Jamil, Hongsu Byun, and Youngjae Kim's paper on "KVAccel: A Novel Write Accelerator for LSM-Tree-Based KV Stores with Host-SSD Collaboration" has been accepted in International Parallel and Distributed Processing Symposium (IPDPS).

Kyuli Park, Dongjae Lee, Yeonwoo Jeong, Sungyong Park's paper titled "Towards an Asynchronous Commit in Micro-batch Streaming Systems with Log-Structured Merge-tree based Key-Value Store" has been accepted in Cluster Computing Journal (SCIE).

Research in the field of AI Systems: Kihyun Kim, and Youngjae Kim's paper on "DRAM 기반 스토리지를 활용한 RAG 기반 LLM 추론 가속화 연구" has been accepted in KCS 2025.

Research in the field of AI Systems: Hyungwoo Lee, Kihyun Kim, and Youngjae Kim's paper on "High-Performance LLM RAG System Utilizing Disk-Based KV Cache in Vector Database" has been accepted in KSC 2024.

Research in the field of AI Systems: Jinwoo Kim, Kihyun Kim, Yoochan Kim and Youngjae Kim's paper on "Resolving I/O Bottlenecks through Partition Prediction and Preloading in Large-Scale Disk-based GNN Training" has been accepted in KSC 2024.

Safdar Jamil, Joseph Ro and Youngjae Kim's paper on "Efficient Data Placement in Deduplication enabled ZenFS via CRC-Based Prediction" has been accepted in IEEE Access.

GyeongHwan Jung, Yeonwoo Jeong, Kyuri Park, Dongjae Lee, Hongsu Byun, Suyeon Lee, Sungyong Park's paper on "dStream: An Online-based Dynamic Operator-level Query Mapping Scheme on Discrete CPU-GPU Architectures" has been accepted in IEEE Access.

Soon Hwang, Junhyeok Park, Junghyun Ryu, Youngjae Kim's WIP Proposal titled "Improving SQL Query Execution of a Distributed SQL Query Engine on Object-Based Computational Storage through Multi-Layered Offloading" has been accepted for presentation in the 9th International Parallel Data Systems Workshop (PDSW'24).

Our DISCOS Lab. received the Mid-career Research Grant (중견연구) for “Research on Hardware-Software Co-Design for Building a Software Defined Computational Storage Ecosystem” (National Research Foundation of Korea, 2024.09 ~ 2029.08) (PI: Prof. Sungyong Park, Co-PI: Prof. Yongjae Kim).

Junghyun Ryu was selected for the Master's Student Research Encouragement Grant from the National Research Foundation of Korea. He will conduct research over the next year on the topic of exploring the utility of DPU in storage systems. Congratulations!

Seoyeong Lee, Awais Khan, Yoochan Kim, Junghwan Park, Soon Hwang, Youngjae Kim’s paper titled ‘Fault-Tolerant Deep Learning Cache with Hash Ring for Load Balancing in HPC Systems’ has been accepted in Proceedings of the 9th International Parallel Data Systems Workshop (PDSW’24).

Farid Talibli has joined us as a new MS/PhD student. Welcome aboard!

DISCOS Lab received new funding (2024) from NRF under the research title "대규모 머신러닝 가속화를 위한 고성능 계산-데이터 플랫폼 연구", 서강대학교, 2024년도 과학기술정보통신부 글로벌 기초연구실 지원사업 신규과제 선정 (연구책임자: 김영재 교수, 공동연구원: 박성용 교수).

Honghyeon Yoo, Hongsu Byun and Sungyong Park’s poster titled ‘Refining Compaction Offloading I/O Stack for LSM-based Key-Value Stores with SPDK’ has been accepted in the IEEE International Conference on Cluster Computing (IEEE CLUSTER’24).

Hamin Hwangbo, Joseph Ro, Sungjin Byeon, Safdar Jamil and Youngjae Kim's paper on "Towards A Unified Garbage Collection Strategy in ZNS Key-Value Store File Systems Using Same-Victim GC" has been accepted for presentation at the 32nd International Symposium on the Modeling, Analysis, and Simulation of Computer and Telecommunication Systems (MASCOTS).

Yonghyeon Cho, Yoochan Kim, Kihyun Kim, Jinwoo Kim, Youngjae Kim's paper on "Optimizing Multi-level Checkpointing for Distributed Deep Learning Workloads on Cloud Spot VM Clusters" has been accepted in IEEE Access.

The paper titled 'ML-based Dynamic Operator-Level Query Mapping for Stream Processing Systems in Heterogeneous Computing Environments' authored by Sejeong Oh, Gordon Euhyun Moon and Sungyong Park, has been accepted in the IEEE International Conference on Cluster Computing (IEEE CLUSTER`24).

Kihwan Kim, and Jaewan Park have joined us as new MS students. Welcome on board!

Junghyun Ryu, Hongsu Byun, Youngjae Kim's poster titled "Evaluation of Erasure Coding and Opportunistic Offloading Algorithms using DPU in Distributed Storage Systems" has been accepted in the IEEE Non-Volatile Memory Systems and Applications Symposium (NVMSA).

Dr. Awais Khan (DISCOS alumnus - Advisor: Prof. Youngjae Kim, currently HPC Systems Scientist at Oak Ridge National Laboratory)'s paper - "An Evaluation of the Effect of Network Cost Optimization for Leadership Class Supercomputers" submitted to SC 2024 received Best Paper Nomination. All DISCOS members wholeheartedly congratulate Dr. Awais Khan.

The paper titled 'BandSlim: A Novel Bandwidth and Space-Efficient KV-SSD with an Escape-from-Block Approach' authored by Junhyeok Park, Chang-Gyu Lee, Soon Hwang, and Youngjae Kim, has been accepted for presentation at the 53rd International Conference on Parallel Processing (ICPP).

Michel Nguyen, Yazhu Song (Florida State University) and Beiyu Wang (University of Florida) have joined DISCOS as visiting researchers for 11 weeks. Welcome on board!

The paper titled 'Coordinating Compaction between LSM-tree based Key-Value Stores for Edge Federation' authored by Jeeseob Kim, Honghyeon Yoo, Seungjae Lee, Hongsu Byun and Sungyong Park, has been accepted in the IEEE International Conference on Cloud Computing
(IEEE CLOUD'24).

Soon (PhD student) completed an internship at Oak Ridge National Lab during the last winter vacation, and as a result, contributed to the publication of an I/O acceleration poster for distributed deep learning at SOS26.

The paper titled 'Ensuring Compaction and Zone Cleaning Efficiency through Same-Zone Compaction in ZNS Key-Value Store,' authored by Sungjin Byeon, Joseph Ro, and Youngjae Kim, has been accepted for presentation at the 38th International Conference on Massive Storage Systems and Technology (MSST).

Hyungwoo Lee has joined us as a new MS student. Welcome aboard!

The paper titled 'DeepVM: Integrating Spot and On-Demand VMs for Cost-Efficient Deep Learning Clusters in the Cloud,' authored by Yoochan Kim, Kihyun Kim, Jinwoo Kim, Youngjae Kim, has been accepted for presentation at the 24th IEEE/ACM International Symposium on Cluster, Cloud and Internet Computing (CCGrid).

Kyuri Park, Sungyong Park's paper titled "Performance Improvement Using Multi Checkpointing Technique in Micro-batch Streaming System" has been selected as the best paper presentation award in the field of the high performance computing at Korea Software Congress (KSC'23).

The paper titled 'OctoFAS: A Two-Level Fair Scheduler for Enhancing Fairness in Network-based Key-Value Storage,' authored by Yeohyeon Park, Junhyeok Park, Junghwan Park, Awais Khan, Youngjae Kim, has been accepted in the MDPI Electronics (Electronics).

Soyang Baek, Dongjae Lee, Honghyeon Yoo, Sungjin Byeon and Junhyeok Park have joined us as new MS students. Welcome on board!