SSDUPplus
The new version of SSDUP, an optimized SSD Burst Buffer for HPC by Traffic Detection
Install / Use
/learn @CGCL-codes/SSDUPplusREADME
SSDUP+
Introduction
SSDUP+ is a traffic-aware optimazation for Burst Buffer, which effectively solves the burden of Burst Buffer on large SSD capacity and overlapping computation with data flushing. The prototype of SSDUP+ is based on OrangeFS (http://www.orangefs.org/).
How to use
Start
orangefs_install/sbin/pvfs2-server orangefs-server.conf wb_info.conf -f
orangefs_install/sbin/pvfs2-server orangefs-server.conf wb_info.conf
Sample wb_info.conf
buff_path_1 : /mnt/ssd1
buff_path_2 : /mnt/ssd2
buff_threshold : 8GB
high_watermark : 0.45
low_watermark : 0.3
rs_length : 128
Compile
chage src/io/trove/trove-handle-mgmt/module.mk to:
DIR := src/io/trove/trove-handle-mgmt
SERVERSRC += \
$(DIR)/avltree.c \
$(DIR)/my_avltree.c \
$(DIR)/trove-extentlist.c \
$(DIR)/trove-ledger.c \
$(DIR)/trove-handle-mgmt.c
Other operations remain the same as orangefs
Reference Paper
Xuanhua Shi, Ming Li, Wei Liu, Hai Jin, Chen Yu, and Yong Chen, "SSDUP: A Traffic-Aware SSD Burst Buffer for HPC Systems". in Proceedings of the ACM International Conference on Supercomputing (ICS), Chicago, Illinois, USA, June 13-16, 2017.
Xuanhua Shi, Wei Liu, Ligang He, Hai Jin, Ming Li, and Yong Chen, "Optimizing the SSD Burst Buffer by Traffic Detection", ACM Transactions on Architecture and Code Optimization, Vol. 17, No. 1, Article 8, March 2020, 26 pages.
Support Or Contact
If you have any questions, please contact Wei Liu (cccloude@hust.edu.cn)
Related Skills
node-connect
352.5kDiagnose OpenClaw node connection and pairing failures for Android, iOS, and macOS companion apps
frontend-design
111.3kCreate distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, or applications. Generates creative, polished code that avoids generic AI aesthetics.
openai-whisper-api
352.5kTranscribe audio via OpenAI Audio Transcriptions API (Whisper).
qqbot-media
352.5kQQBot 富媒体收发能力。使用 <qqmedia> 标签,系统根据文件扩展名自动识别类型(图片/语音/视频/文件)。
