Seoyoung Ko
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Seoyoung Ko has a total of 24 co-authors. Here are the top co-authors by number of shared publications:
- Jung Ho Ahn
Seoul National University, Seoul, Republic of Korea
5 shared publications - Wanju Doh
Seoul National University, Seoul, Republic of Korea
4 shared publications - Eojin Lee
Inha University, Incheon, Republic of Korea
3 shared publications - Hyunjeong Shim
Seoul National University, Seoul, Republic of Korea
2 shared publications - Jinin So
Samsung Electronics, Seoul, Korea
2 shared publications - Michael Jaemin Kim
Seoul National University, Seoul, South Korea
2 shared publications
Latest Publications
- Open Access
- 5 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 795-810
Clone: A Collaborative Multi-device System for Retrieval-Augmented Generation over CXL
- 5 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 795-810
Open AccessClone: A Collaborative Multi-device System for Retrieval-Augmented Generation over CXL
- 14 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Computer Architecture Letters
- Vol. 24 (1) , 173-176
Cosmos: A CXL-Based Full In-Memory System for Approximate Nearest Neighbor Search
Retrieval-Augmented Generation (RAG) is crucial for improving the quality of large language models by injecting proper contexts extracted from external sources. RAG requires high-throughput, low-latency Approximate Nearest Neighbor Search (ANNS) over billion-scale vector databases. Conventional DRAM/SSD solutions face capacity/latency limits, whereas specialized hardware or RDMA clusters lack flexibility or incur network overhead. We present COSMOS, integrating general-purpose cores within CXL memory devices for full ANNS offload and introducing rank-level parallel distance computation to maximize memory bandwidth. We also propose an adjacency-aware data placement that balances search loads across CXL devices based on inter-cluster proximity. Evaluations on SIFT1B and DEEP1B traces show that COSMOS achieves up to 6.72× higher throughput than the baseline CXL system and 2.35× over a state-of-the-art CXL-based solution, demonstrating scalability for RAG pipelines.
- 14 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Computer Architecture Letters
- Vol. 24 (1) , 173-176
Cosmos: A CXL-Based Full In-Memory System for Approximate Nearest Neighbor Search
Retrieval-Augmented Generation (RAG) is crucial for improving the quality of large language models by injecting proper contexts extracted from external sources. RAG requires high-throughput, low-latency Approximate Nearest Neighbor Search (ANNS) over billion-scale vector databases. Conventional DRAM/SSD solutions face capacity/latency limits, whereas specialized hardware or RDMA clusters lack flexibility or incur network overhead. We present COSMOS, integrating general-purpose cores within CXL memory devices for full ANNS offload and introducing rank-level parallel distance computation to maximize memory bandwidth. We also propose an adjacency-aware data placement that balances search loads across CXL devices based on inter-cluster proximity. Evaluations on SIFT1B and DEEP1B traces show that COSMOS achieves up to 6.72× higher throughput than the baseline CXL system and 2.35× over a state-of-the-art CXL-based solution, demonstrating scalability for RAG pipelines.
- Open Access
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 854-869
PET: Proactive Demotion for Efficient Tiered Memory Management
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 854-869
Open AccessPET: Proactive Demotion for Efficient Tiered Memory Management
- 11 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Embedded Systems Letters
- Vol. 17 (3) , 192-195
Hechi: A Hybrid Approach for Efficient Memory Reclamation Techniques in Mobile Systems
Application startup time, the time it takes for an application to become visible to the user from startup, is a crucial factor affecting user experience in mobile systems. The startup time of applications switching from the background (switching time) depends largely on the number of pages read from storage. When an application starts for the first time or reloads after being killed, it suffers from longer startup time (cold-launch time) as it begins from scratch. To mitigate this, minimizing application kills while retaining essential data in main memory can help, but it involves a trade-off: reclaiming more file-backed pages (pages) slows down switching due to the overhead of fetching them from storage, while reclaiming more anonymous pages (pages) that not associated with any file risks app termination due to limited swap space. addresses this trade-off using a hybrid memory reclamation approach. By deprioritizing the reclamation of essential file pages and dynamically adjusting the reclamation ratio of the two types of pages based on system states, improves startup time by up to 17%.
- 11 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Embedded Systems Letters
- Vol. 17 (3) , 192-195
Hechi: A Hybrid Approach for Efficient Memory Reclamation Techniques in Mobile Systems
Application startup time, the time it takes for an application to become visible to the user from startup, is a crucial factor affecting user experience in mobile systems. The startup time of applications switching from the background (switching time) depends largely on the number of pages read from storage. When an application starts for the first time or reloads after being killed, it suffers from longer startup time (cold-launch time) as it begins from scratch. To mitigate this, minimizing application kills while retaining essential data in main memory can help, but it involves a trade-off: reclaiming more file-backed pages (pages) slows down switching due to the overhead of fetching them from storage, while reclaiming more anonymous pages (pages) that not associated with any file risks app termination due to limited swap space. addresses this trade-off using a hybrid memory reclamation approach. By deprioritizing the reclamation of essential file pages and dynamically adjusting the reclamation ratio of the two types of pages based on system states, improves startup time by up to 17%.
- Open Access
- 28 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 986-1001
How to Kill the Second Bird with One ECC: The Pursuit of Row Hammer Resilient DRAM
Error-correcting code (ECC) has been widely used in DRAM-based memory systems to address the exacerbating random errors following the fabrication process scaling. However, ECCs including the strong form of Chipkill have not been so effective against Row Hammer (RH), which incurs bursts of errors discretely corrupting the whole row beyond the ECC correction capability. We propose Cube, a novel chip-wise physical to DRAM address randomization scheme that leverages the abundant detection capability of on-die-ECC (OECC) and correction capability of Chipkill against RH. Cube allows for synergistic cooperation between ECC, probabilistic RH-protection schemes, and the system with minimal to no modification for each. First, Cube scrambles the rows of each chip using a boot-time key in a way that distributes RH victims to multiple Chipkill codewords. Second, Cube utilizes the newly observed distinct RH error characteristics from real DRAM chips, to swiftly diagnose the RH victim rows using the error profile from OECC scrubbing, and even correct it leveraging Chipkill. When combined, Cube decreases the failure probability of PARA and a state-of-the-art RH protection scheme SRS by up to 10− 25. At a target failure probability of 10− 10 per year on a DDR5 rank under the RH threshold of 2K, Cube reduces the performance and table size overhead of SRS by up to 24.3% and 39.9%, respectively.
- 28 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 986-1001
Open AccessHow to Kill the Second Bird with One ECC: The Pursuit of Row Hammer Resilient DRAM
Error-correcting code (ECC) has been widely used in DRAM-based memory systems to address the exacerbating random errors following the fabrication process scaling. However, ECCs including the strong form of Chipkill have not been so effective against Row Hammer (RH), which incurs bursts of errors discretely corrupting the whole row beyond the ECC correction capability. We propose Cube, a novel chip-wise physical to DRAM address randomization scheme that leverages the abundant detection capability of on-die-ECC (OECC) and correction capability of Chipkill against RH. Cube allows for synergistic cooperation between ECC, probabilistic RH-protection schemes, and the system with minimal to no modification for each. First, Cube scrambles the rows of each chip using a boot-time key in a way that distributes RH victims to multiple Chipkill codewords. Second, Cube utilizes the newly observed distinct RH error characteristics from real DRAM chips, to swiftly diagnose the RH victim rows using the error profile from OECC scrubbing, and even correct it leveraging Chipkill. When combined, Cube decreases the failure probability of PARA and a state-of-the-art RH protection scheme SRS by up to 10− 25. At a target failure probability of 10− 10 per year on a DDR5 rank under the RH threshold of 2K, Cube reduces the performance and table size overhead of SRS by up to 24.3% and 39.9%, respectively.