전자통신동향분석 (Electronics and Telecommunications Trends)

한국전자통신연구원 (Electronics and Telecommunications Research Institute)
권호 표지
DOI QR코드

DOI QR Code

CXL 인터커넥트 기술 연구개발 동향

Trends in Compute Express Link(CXL) Technology

  • 김선영 (슈퍼컴퓨팅기술연구센터) ;
  • 안후영 (슈퍼컴퓨팅기술연구센터) ;
  • 박유미 (슈퍼컴퓨팅기술연구센터) ;
  • 한우종 (슈퍼컴퓨팅기술연구센터)
  • S.Y. Kim ;
  • H.Y. Ahn ;
  • Y.M. Park ;
  • W.J. Han
  • 발행 : 2023.10.01
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

With the widespread demand from data-intensive tasks such as machine learning and large-scale databases, the amount of data processed in modern computing systems is increasing exponentially. Such data-intensive tasks require large amounts of memory to rapidly process and analyze massive data. However, existing computing system architectures face challenges when building large-scale memory owing to various structural issues such as CPU specifications. Moreover, large-scale memory may cause problems including memory overprovisioning. The Compute Express Link (CXL) allows computing nodes to use large amounts of memory while mitigating related problems. Hence, CXL is attracting great attention in industry and academia. We describe the overarching concepts underlying CXL and explore recent research trends in this technology.

키워드

과제정보

본 연구 논문은 한국연구재단 슈퍼컴퓨터개발선도사업[2021M3H6A1017683, 초병렬프로세서 기반 고집적 컴퓨팅 노드 및 시스템 개발]의 일환으로 수행되었음.

참고문헌

  1. T. Brown et al., "Language models are Few-shot learners," in Proc. NeurIPS 2020, (Vancouver, Canada), 용어해설 Dec. 2020, pp. 1877-1901.
  2. K.S. Kim et al., "SMT: Software-defined memory tiering for heterogeneous computing systems with cxl memory expander," IEEE Micro, vol. 43, no. 2, 2023, pp. 20-29.
  3. H. Li et al., "Pond: CXL-based memory pooling systems for cloud platforms," in Proc. ACM ASPLOS 2023, vol. 2, (Vancouver, Canada), Jan. 2023.
  4. 김선영 외, "CCIX 연결망과 메모리 확장기술 동향," 전자통신동향분석, 제37권 제1호, 2022, pp. 42-52.
  5. 오명훈 외, "메모리 중심 컴퓨팅 기술 동향," 융합연구리뷰, vol. 6, no. 3, 2020.
  6. CXL Consortium, CXL 3.0 Specification, https://www.computeexpresslink.org/download-the-specification
  7. Compute Express Link 3.0, https://www.computeexpresslink.org/_files/ugd/0c1418_a8713008916044ae9604405d10a7773b.pdf
  8. CCIX, About the CCIX Consortium, https://www.ccixconsortium.com/about/
  9. Tektronix, "PAM4 Signaling in High-Speed Serial Technology: Test, Analysis, and Debug," Available: https://download.tek.com/document/PAM4-Signalingin-High-Speed-Serial-Technology_55W-60273.pdf
  10. A. Biswas, "Sapphire rapids," in Proc. IEEE HCS 2021, (Palo Alto, CA, USA), Aug. 2021, pp. 1-22.
  11. AMD EPYCTM 9004 Series Server Processors Launch, https://www.amd.com/en/events/epyc4
  12. S.J. Park et al., "Scaling of memory performance and capacity with CXL memory expander," in Proc. IEEE HCS 2022, (Palo Alto, CA, USA), Aug. 2022, pp. 1-27.
  13. SK hynix Newsroom, "SK hynix Develops DDR5 DRAM CXLTM Memory to Expand the CXL Memory Ecosystem," 2022, https://news.skhynix.com/skhynix-develops-ddr5-dram-cxltm-memory-toexpand-the-cxl-memory-ecosystem/
  14. CZ120 memory expansion module, https://www.micron.com/solutions/server/cxl
  15. Scalable Memory Development Kit, https://github.com/OpenMPDK/SMDK
  16. Astera Labs, "Leo Memory Connectivity Platform for CXL 1.1 and 2.0," 2022, https://www.asteralabs.com/products/cxl-memory-platform/leo-cxl-memoryconnectivity-platform/
  17. XC50256 CXL2.0/PCle5.0 switch, https://www.xconntech.com/product
  18. Enfabrica at ISC 2023, Scaling CXL Memory Using High Speed Networking, https://www.youtube.com/watch?v=YdJWhqeT5DM
  19. M.H. Ha et al., "Dynamic capacity service for improving CXL pooled memory efficiency," IEEE Micro, vol. 43, no. 2, 2023, pp. 39-47. https://doi.org/10.1109/MM.2023.3237756
  20. D.H. Gouk et al., "Direct access, high performance memory disaggregation with DirectCXL," in Proc. USENIX ATC 2022, (Carlsbad, CA, USA), Jul. 2022, pp. 287-294.
  21. D.S. Berger et al., "Design tradeoffs in CXL-based memory pools for public cloud platforms," IEEE Micro, vol. 43, no. 2, 2023, pp. 30-38. https://doi.org/10.1109/MM.2023.3241586
  22. W. Huangfu et al., "BEACON: Scalable near data processing accelerators for genome analysis near memory pool with the CXL support," in Proc. IEEE/ACM MICRO 2022, (Chicago, IL, USA), Oct. 2022, pp. 727-743.
  23. Y.G. Kim et al., "Ramulator: A fast and extensible DRAM simulator," IEEE Comput. Archit. Lett., vol. 15, no. 1, 2016, pp. 45-49. https://doi.org/10.1109/LCA.2015.2414456
  24. J.S. Sim et al., "Computational CXL-memory solution for accelerating memory-intensive applications," IEEE Comput. Archit. Lett., vol. 22, no. 1, 2023, pp. 5-8. https://doi.org/10.1109/LCA.2022.3226482
  25. KNN benchmark, https://github.com/mnms/ltdb-knnkernel-bench
  26. Y. Yang et al., "CXLMemSim: A pure software simulated CXL.mem for performance characterization," arXiv preprint, CoRR, 2023, arXiv: 2303.06153.
  27. CXLMemSim, https://github.com/SlugLab/CXLMemSim
  28. gem5-cxl, https://github.com/fadedzipper/gem5-cxl