Electronics and Telecommunications Trends (전자통신동향분석)

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

DOI QR Code

Technology Trends in CXL Memory and Utilization Software

CXL 메모리 및 활용 소프트웨어 기술 동향

  • H.Y. Ahn ;
  • S.Y. Kim ;
  • Y.M. Park ;
  • W.J. Han
  • 안후영 (슈퍼컴퓨팅기술연구센터 ) ;
  • 김선영 (슈퍼컴퓨팅기술연구센터) ;
  • 박유미 (슈퍼컴퓨팅기술연구센터 ) ;
  • 한우종 (슈퍼컴퓨팅기술연구센터 )
  • Published : 2024.02.01
Download PDF
⟨ Previous Next ⟩

Abstract

Artificial intelligence relies on data-driven analysis, and the data processing performance strongly depends on factors such as memory capacity, bandwidth, and latency. Fast and large-capacity memory can be achieved by composing numerous high-performance memory units connected via high-performance interconnects, such as Compute Express Link (CXL). CXL is designed to enable efficient communication between central processing units, memory, accelerators, storage, and other computing resources. By adopting CXL, a composable computing architecture can be implemented, enabling flexible server resource configuration using a pool of computing resources. Thus, manufacturers are actively developing hardware and software solutions to support CXL. We present a survey of the latest software for CXL memory utilization and the most recent CXL memory emulation software. The former supports efficient use of CXL memory, and the latter offers a development environment that allows developers to optimize their software for the hardware architecture before commercial release of CXL memory devices. Furthermore, we review key technologies for improving the performance of both the CXL memory pool and CXL-based composable computing architecture along with various use cases.

Keywords

Acknowledgement

본 논문은 한국전자통신연구원 내부연구개발사업 대규모 데이터 처리 응용 성능 향상을 위한 새로운 메모리 연결망(CXL) 기술 기반 MPI 분산병렬처리 구조 연구 및 성능분석(23YS1700)의 지원을 받아 수행된 연구임.

References

  1. J. Hoffmann et al., "Training compute-optimal large language models," arXiv preprint, CoRR, 2022, arXiv: 2203.15556. 
  2. J.A. Baktash and M. Dawodi, "Gpt-4: A review on advancements and opportunities in natural language processing," arXiv preprint, CoRR, 2023, arXiv: 2305.03195. 
  3. C. Guo et al., "Exploring the benefits of resource disaggregation for service reliability in data centers," IEEE Trans. on Cloud Comput., vol. 11, no. 2, 2022, pp. 1651-1666. 
  4. I.H. Chung, B. Abali, and P. Crumley, "Towards a composable computer system," in Proc. HPC Asia 2018, (Tokyo Japan), Jan. 2018, pp. 137-147. 
  5. MemVerge: The Road to Endless Memory, https://www.youtube.com/watch?v=rO4PdTAwLTY&t=622s 
  6. [CES 2023 Innovation Award] Publishing the Boundaries of Memory, https://semiconductor.samsung.com/emea/news-events/tech-blog/pushing-the-boundaries-of-memory-samsung-goes-beyond-hardware-to-become-total-solution-provider-with-cxl-memory-expander/ 
  7. SK hynix CXL Memory, https://news.skhynix.com/sk-hynix-develops-ddr5-dram-cxltm-memory-to-expand-the-cxl-memory-ecosystem/ 
  8. Enfabrica: Scaling CXL Memory Using High Speed Networking, https://www.youtube.com/watch?v=YdJWhqeT5DM&t=919s 
  9. Xconn CXL Switches: Enablers of More Advanced HPC and AI/ML Cloud Computing, https://www.youtube.com/watch?v=VvKEHq3xjUw&list=PLsf8NUp2sz_iF3X4s_qazc8c_vYnlAVXv&index=4 
  10. J. Sim et al., "Computational CXL-memory solution for accelerating memory-intensive applications," IEEE Comput. Archit. Lett., vol. 22, no. 1, 2022, pp. 5-8. 
  11. Y. Fridman et al., "CXL Memory as Persistent Memory for disaggregated HPC: A practical approach," arXiv preprint, CoRR, 2023, arXiv: 2308.10714. 
  12. D. Lee et al., "Elastic use of far memory for In-memory database management systems," in Proc. DaMoN 2023, (Seattle, WA, USA), June 2023, pp. 35-43. 
  13. Y. Sun et al., "Demystifying CXL memory with genuine CXL-ready systems and devices," in Proc. MICRO 2023, (Toronto, Canada), Oct. 2023, pp. 105-121. 
  14. K. 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. 
  15. H. Li et al., "Pond: CXL-based memory pooling systems for cloud platforms," in Proc. ASPLOS 2023, vol. 2, (Vancouver, Canada), Mar. 2023, pp. 574-587. 
  16. D. Gouk et al., "Memory pooling with cxl," IEEE Micro, vol. 43, no. 2, 2023, pp. 48-57. 
  17. M. Kwon et al., "Failure tolerant training with persistent memory disaggregation over CXL," IEEE Micro, vol. 43, no. 2, 2023, pp. 66-75. 
  18. Flight Simulator, https://memverge.com/memverge-and-sk-hynix-accelerate-memory-pooling-and-sharing-software-development-with-cxl-flight-simulator/ 
  19. M. Ha et al., "Dynamic capacity service for improving CXL pooled memory efficiency," IEEE Micro, vol. 43, no. 2, 2023, pp. 39-47. 
  20. Jonathan Cameron's QEMU Working Fork, https://gitlab.com/jic23/qemu 
  21. Y. Yang et al., "CXLMemSim: A pure software simulated CXL. mem for performance characterization," arXiv preprint, CoRR, 2023, arXiv: 2303.06153. 
  22. Scalable Memory Development Kit(SMDK) v1.5, https://github.com/OpenMPDK/SMDK 
  23. HMSDK, https://github.com/skhynix/hmsdk 
  24. MemVerge Project Gismo: Global IO-free Shared Memory Objects, https://www.youtube.com/watch?v=D66W7eqFbhc 
  25. Ray, https://www.ray.io/ray-sgd 
  26. Timeplus, https://www.timeplus.com/ 
  27. MemVerge Memory Viewer CXL and Memory Machine CXL, https://www.youtube.com/watch?v=cwZCORGjCsM 
  28. GII Korea, "세계의 CXL 컨트롤러 IP 시장 분석," 2023, https://www.giikorea.co.kr/report/qyr1215847-global-cxl-controller-ip-market-research-report.html