Date: Thu, 03 May 2012 10:46:12 +0900
--------------------------------------------------
2012年 4月 23日~25日に開催された Lustre User Group (LUG) 2012の
スライドとビデオが公開されています:
LUG 2012 Agenda
http://www.opensfs.org/lug/program
Presentations
http://www.opensfs.org/resources-2/presentations
Media (YouTubeへのリンク)
http://www.opensfs.org/resources-2/videos
個人的に気になったシステム関連の発表:
==========
"Lustre Future Development"
Andreas Dilger, Whamcloud
http://www.opensfs.org/wp-content/uploads/2011/11/LUG-2012-Lustre-future-development.pdf
Lustre 2.4 - March 2013
Lustre 2.4 - ZFS OSD (WC/LLNL) (Page 6)
Leverage many features immediately
Data checksums on disk + Lustre checksums on network
Online filesystem check/scrub/repair- no more e2fsck!
Integrated with flash storage cache (L2ARC)
More ZFS features to leverage in the future
Snapshots, end-to-end data integrity, datasets
Lustre 2.4 - HSM (WC/CEA) (Page 7)
Originally developed by CEA France
Simple archive back-end interface
Infrastructure useful for other projects
ZFS on Linux Licensing Answers (Page 13)
==========
"Sequoia's 55PB Lustre+ZFS Filesystem"
Brian Behlendorf, Lawrence Livermore National Laboratory
http://www.opensfs.org/wp-content/uploads/2011/11/LUG12_ZFS_Lustre_for_Sequoia.pdf
Status Update (Page 5)
55PB Lustre+ZFS File system for Sequoia
Development still under way
Performance work started
Contract with Whamcloud
MDS Configurations (Page 12)
LDISKFS+MD & ZFS+MIRROR
OSS Configurations (Page 16)
LDISKFS+RAID6 & ZFS+RAID6 & ZFS+RAIDZ2
Website
http://zfsonlinux.org
http://zfsonlinux.org/lustre.html
※LDISKFSと ZFSとの比較等、いろいろ試行錯誤の感じがします。
==========
"Lustre 2.1 and Lustre-HSM at IFERC"
Diego Moreno, Bull
http://www.opensfs.org/wp-content/uploads/2011/11/LUG_2012_Diego_Moreno_v2.pdf
※六ヶ所村の「六ちゃん」
Archive Lustre L2 - DMF (HSM) (Page 7, 8)
8PB Lustre + 40PB SGI DMF (Lustre Clients)
==========
"Current Status of FEFS for the K Computer"
Shinji Sumimoto, Fujitsu
http://www.opensfs.org/wp-content/uploads/2011/11/LUG2012-FJ-20120426.pdf
==========
"A Technical Overview of the OLCF's Next-generation Center-wide Lustre File System"
Sarp Oral, Oak Ridge National Laboratory
http://www.opensfs.org/wp-content/uploads/2011/11/oral-lug-2012.pdf
Transitioning I/O to next gen computing (Page 2)
From Jaguar to Titan
Number of cores: 224K -> 300K
Memory: 300 TB -> 600 TB
Peak Performance: 2.2 PFlops -> 10-20 Pflops
Proprietary Interconnect: SeaStar2+ -> Gemini
Peak egress I/O (over IB): (192 x 1.5 GB/s) -> (384-420 x 2.8-3 GB/s)
New Architecture (Page 5)
Target numbers for next gen parallel file system
1 TB/s file system-level well-formed I/O performance
240 GB/s file system-level random I/O performance
Capacity will be based on the selected storage media
Expected to be 9-20 PB
Availability: >95%
Expected availability will be similar of Spider's
0 件のコメント:
コメントを投稿