DOI QR코드

DOI QR Code

Performance Enhancement Method Through Science DMZ Data Transfer Node Tuning Parameters

Science DMZ 데이터 전송 노드 튜닝 요소를 통한 성능 향상 방안

  • 박종선 (한국과학기술정보연구원 첨단연구망서비스실) ;
  • 박진형 (한국과학기술정보연구원 정보시스템운영실) ;
  • 김승해 (한국과학기술정보연구원 첨단연구망서비스실) ;
  • 노민기 (한국과학기술정보연구원 첨단연구망서비스실)
  • Received : 2018.01.22
  • Accepted : 2018.01.30
  • Published : 2018.02.28

Abstract

In an environment with a large network bandwidth, maximizing bandwidth utilization is an important issue to increase transmission efficiency. End-to-end transfer efficiency is significantly influenced by factors such as network, data transfer nodes, and intranet network security policies. Science DMZ is an innovative network architecture that maximizes transfer performance through optimal solution of these complex components. Among these, the data transfer node is a key factor that greatly affects the transfer performance depending on storage, network interface, operating system, and transfer application tool. However, tuning parameters constituting a data transfer node must be performed to provide high transfer efficiency. In this paper, we propose a method to enhance performance through tuning parameters of 100Gbps data transfer node. With experiment result, we confirmed that the transmission efficiency can be improved greatly in 100Gbps network environment through the tuning of Jumbo frame and CPU governor. The network performance test through Iperf showed improvement of 300% compared to the default state and NVMe SSD showed 140% performance improvement compared to hard disk.

네트워크 대역폭이 큰 환경에서는 대역폭 활용률을 극대화함으로써 전송효율성을 높이는 것이 매우 중요한 이슈이다. 종단간 전송효율성은 네트워크, 데이터 전송 노드 그리고 기관 내 네트워크 보안정책 등 구성요소에 따라 크게 영향을 받는다. Science DMZ는 이러한 복합적인 구성요소들의 최적의 해결 방안을 통해 전송성능을 극대화하기 위한 혁신적인 네트워크 구조이다. 이 중에서 데이터 전송 노드는 스토리지, 네트워크 인터페이스, 운영체제, 전송응용 도구에 따라 전송성능에 크게 영향을 주는 핵심 요소이다. 하지만 고속네트워크 환경에서는 데이터 전송 노드를 구성하는 요소들의 적절한 튜닝이 수행되어야 높은 전송효율성을 제공할 수 있다. 본 논문에서는 100Gbps 데이터 전송 노드의 튜닝 요소를 통한 전송성능 향상 방안에 대해 제안한다. 성능측정결과 점보프레임, CPU governor 튜닝을 통해 100Gbps 네트워크 환경에서 전송효율성을 크게 개선할 수 있음을 확인하였다. Iperf를 통한 네트워크 성능테스트 결과 default에 비해 300%의 성능향상을 보였으며 NVMe SSD의 경우 하드디스크와 비교해 140%의 성능개선을 확인하였다.

Keywords

References

  1. D. Eli, R. Lauren, T. Brian, H. Mary, and Z. Jason, "The Science DMZ: A network design pattern for data-intensive science," in Proceeding of the International Conference on High Performance Computing, Networking, Storage and Analysis, 2013.
  2. V. Nagendra, V. Yegneswaran, and P. Porras, "Securing ultra-high-bandwidth science DMZ networks with coordinated situational awareness," in Proceedings of the 16th ACM Workshop on Hot Topics in Networks, 2017.
  3. S. Kim and J. Yang, "Work-in-progress: improving NVMe SSD I/O determinism with PCIe virtual channel," in Proceeding of International Conference on Compilers, Architectures and Synthesis For Embedded Systems, 2017.
  4. J. Qian, H. Jiang, S. A. Witawas, S. Seth, S. Skelton, and J. Moore, "Energy-Efficient I/O Thread Schedulers for NVMe SSDs on NUMA," in Proceeding of International Symposium on Cluster, Cloud and Grid Computing, 2017.
  5. Q. Xu et al., "Performance analysis of NVMe SSDs and their implication on real world databases," in Proceedings of the 8th ACM International Systems and Storage Conference, 2015.
  6. K. Mesmin and J. Mbyamm, J. Zhang, "Improved implementation of TCP-vegas method in interchanges of satellite links," in Proceeding of International Conference on Computer Science and Network Technology, 2016.
  7. J. Cheng, D. X. Wei, and S. H. Low, "FAST TCP: motivation, architecture, algorithms, performance," in Proceeding of INFOCOM on IEEE Computer and Communications Societies, 2004.
  8. S. Floyd, "HighSpeed TCP for large congestion windows," IETF, RFC3649, 2003.
  9. Q. Liu, N. Rao, and C. Q. Wu, "Measurement-based performance profiles and dynamics of UDT over dedicated connections," in Proceeding of nternational Conference on Network Protocols, 2016.
  10. E. He, J. Leigh, O. Yu, and T. A. DeFanti, "Reliable Blast UDP : predictable high performance bulk data transfer," in Proceeding of IEEE International Conference on Cluster Computing, 2002.
  11. M. Meiss, "Tsunami: a high-speed rate-controlled protocol for file transfer," www.evl.uic.edu/eric/atpTSUNAMI.pdf, 2009.
  12. Y. Gu, "UDT: a high performance data transport protocol," Ph.D. Thesis, Laboratory for Advanced Computing, Univ. of Illinois at Chicago, 2005.
  13. E. Jung, R. Kettimuthu, and V. Vishwanath, "Toward optimizing disk-to-disk transfer on 100G networks," in Proceeding of 2013 IEEE International Conference on Advanced Networks and Telecommuncations Systems, 2013.
  14. F. Inoue, T. Ito, H. Ohsaki, and M. Imase, "Implementation and evaluation of GridFTP automatic parallelism tuning mechanism for long-fat, networks," in Proceeding of Asia-Pacific Symposium on Information and Telecommunication Technologies, 2008.
  15. J. Park, D. An, and G. Cho, "An adaptive channel number tuning mechanism on parallel transfer with UDT," in Proceeding of 2013 International Conference on Information Networking, 2013.