Next-generation Computing Systems Laboratory

We are pleased to announce that the project titled "Peta-Scale In-Storage Computing for Accelerating Genome Analysis & DNA Sequencing ", led by Prof. Yongtae Kim from our department as the Lead Principal Investigator (Lead-PI), has been officially selected as a Basic Research Laboratory (BRL) for 2024.

Alongside our laboratory, Prof. Inuk Jung and Prof. Myungseok Kim will contribute as Co-Principal Investigators (Co-PIs). This initiative will bring together four research labs within our department, working collaboratively as a research group for up to six years (33 months with a potential 3-year extension).

This project aims to address one of the grand challenges in bioinformatics—the De-novo Assembly problem—by exploring breakthrough solutions from both software and hardware perspectives. We anticipate this research will lay the foundation for Near-Data Processing technology with wide applications across various fields.

Keyword: De-novo Assembly, DNA Sequencing, Computational Biology, Bioinformatics, Near-data processing, Processing-in-Memory, In-storage Computing, Domain-specific Processor

Congratulations! Junghun Lim, a master’s student in our lab, has been selected for the prestigious 2024 IBM Quantum Leadership Training Program.

This program, supported by the Quantum Information Research Support Center and the Ministry of Science and ICT, selects only 15 participants each year from among graduate students, postdoctoral researchers, and industry professionals with over five years of experience across the nation.

Participants will receive four weeks of intensive training in quantum computing fundamentals. During the third week, they will attend an on-site training at the IBM T.J. Watson Research Center in New York, USA.

IBM 2024 Quantum Leadership Training Program: [Link]

Keyword: NISQ Quantum Computing Research, GPU-based Quantum Computing Simulation

This research aims to develop architectures that accelerate bioinformatics applications using Processing-in-Memory (PIM) technology.

Challenges in Software:

• Developing more efficient algorithms for De-novo Assembly

• Creating a programming model for heterogeneous multi-core processing with CPU, GPU, PIM, and ISC

Challenges in Hardware:

• Designing processing elements for bioinformatics applications and optimize them based on their characteristics

• Developing Processing-in-Memory architecture and compiler optimization methods for bioinformatics applications

• Designing In-Storage Processing architecture with offloading techniques suited to bioinformatics workflows

[Related Research Project]

• Peta-Scale In-Storage Computing for Accelerated Genome DNA Sequencing Analysis, Basic Research Laboratory (BRL) Program, National Research Foundation of Korea (2024.08.01 - 2027.04.31)

[Related Fields]

• Computer Architecture, Algorithm, Compiler Design

• Processing-in-Memory

• Bioinformatics/Computational Biology (e.g., DNA read mapping, De-novo Assembly)

As we move beyond Large Language Models (LLMs) toward an era of Artificial General Intelligence (AGI), the demand for vast GPU resources in deep learning model training and inference is becoming increasingly critical. This research aims to alleviate memory bottlenecks in multi-GPU environments by designing a Near-Far Memory Architecture that incorporates (1) far memory based on next-generation non-volatile memory (eNVM) & storage-class memory (SCM) and (2) near memory systems with high-bandwidth, low-latency HBM.

By implementing this scalable Near-Far Memory Architecture in multi-GPU settings, we aim to significantly enhance system performance for LLM operations and in heterogeneous computing environments.

[Related Research Project]

• A Smart Near-Far Memory Architecture for Data-intensive Workloads,
  Yonsei-Samsung Joint Research Program, Samsung Electronics (2021 - )

  • PI: Prof. Bernd Burgstaller, Yonsei University (in joint collaboration)

[Related Fields]

• Computer Architecture, GPGPU Programming, GPU Architecture, CXL Interface

• Heterogeneous Computing, Network-on-Chip (NoC)

• AI for Computer Architecture, Computer Architecture for AI

This research aims to develop high-performance heterogeneous computing systems leveraging xPUs (such as CPUs, GPUs, NPUs, DPUs, QPUs, NDPU, etc.) and multi-tiered memory architectures, including near-far memory and pooled memory with CXL interface. The project encompasses diverse approaches, including the design of heterogeneous computing architectures, management techniques, and the establishment of programming models.

Through this work, we aim to overcome bottlenecks in computation, memory, and network-on-chip (NoC) performance, ultimately re-inventing an HPC system capable of handling a wide range of domains.

[Related Research Projects]

• Regional Intelligence Innovation Talent Development Project (GrandICT/ITRC), funded by IITP (2022 - 2029)

• ICT Convergence Research Center (Lead-PI: Prof. Ji-Hyun Kim): [Link]

• Collaborative Research: Prof. Dukyun Nam (KNU HPC Lab): [Link]

[Related Fields]

• Computer Architecture, Algorithms, Compiler Design

• Programming Model

• Processing-in-Memory, Neural Processing Unit (NPU)