Date: Sun, 28 Nov 2010 06:42:20 +0900
--------------------------------------------------
2010年 6月の NERSCの以下プレスリリースが、2010年 9月の IASDS10で発表
されていました:
"NERSC and HDF Group Optimize HDF5 Library to Improve I/O Performance"
June 28, 2010
http://www.lbl.gov/cs/Archive/news062810.html
"There are several layers of software that deal with I/O on high
performance computing (HPC) systems. The filesystem software, such
as Lustre or GPFS, is closest to the hardware and deals with the
physical access and storage issues. The I/O library, such as HDF5
or netCDF, is closest to the scientific application, and maps the
application's data abstractions onto storage abstractions. I/O
middleware, such as MPI-IO, coordinates activities between the other
layers."
"Getting these three layers of software to work together efficiently
can have a big impact on a scientific code's performance. That's
why the U.S. Department of Energy's (DOE's) National Energy Research
Scientific Computing Center (NERSC) has partnered with the nonprofit
Hierarchical Data Format (HDF) Group to optimize the performance of
the HDF5 library on modern HPC platforms."
発表ペーパー:
"Tuning HDF5 for Lustre File Systems", IASDS10, 2010/09/24
Mark Howison, Quincey Koziol, David Knaak, John Mainzer and John Shalf
http://www.mcs.anl.gov/events/workshops/iasds10/howison_hdf5_lustre_iasds2010.pdf
Agenda,
IASD10 (Workshop on Interfaces and Abstractions for Scientific Data Storage)
http://www.mcs.anl.gov/events/workshops/iasds10/agenda.php
※他にもいろいろな発表があります。
IEEE Cluster 2010
http://www.cluster2010.org/index.html
スライド:
"Tuning HDF5 for Lustre File Systems", IASDS10, 2010/09/24
Mark Howison, Quincey Koziol, David Knaak, John Mainzer and John Shalf
http://www.nersc.gov/projects/presentations/HDF5_DonofrioNERSC.pdf
ペーパーから
Abstract
"We describe our recent work to optimize the performance of the HDF5
and MPI-IO libraries for the Lustre parallel file system. We selected
three different HPC applications to represent the diverse range of
I/O requirements, and measured their performance on three different
systems to demonstrate the robustness of our optimizations across
different file system configurations and to validate our optimization
strategy. We demonstrate that the combined optimizations improve HDF5
parallel I/O performance by up to 33 times in some cases - running
close to the achievable peak performance of the underlying file system
- and demonstrate scalable performance up to 40,960-way concurrency."
検証対象アプリケーション:
*)Global Cloud Resolving Model (GCRM): climate simulation
*)VORPAL: versatile plasma simulation
*)Chombo: adaptive mesh refinement (AMR) package
検証システム
*)JaguarPF: 40,960 Core Cray XT5
*)Franklin: 10.240 Core Cray XT4
*)Hopper (Phase 1): 2560 Core Cray XT5
この成果はいずれ HDF5に取り込まれるそうです (どこかに書いてありました)。
The HDF Group
http://www.hdfgroup.org/
HDF and HDF-EOS Workshop XIV, September 28-30, 2010
AGENDA
http://www.hdfeos.org/workshops/ws14/agenda.php
NERSC関係者の HDF5解説:
"Parallel I/O Tutorial"
Katie Antypas,
Astrosim Summer School on Computational Astrophysics, July 20, 2010
http://supercomputing.astri.umk.pl/files/lectures/Parallelisation.pdf
Schedule, Astrosim Summer School 2010
http://supercomputing.astri.umk.pl/schedule
※ビデオあり
"HDF5 Lab Session: Tutorial from HDF Group" (105 Page)
Katie Antypas,
Astrosim Summer School on Computational Astrophysics, , July 21, 2010
http://www.nersc.gov/projects/presentations/KatieHDF.pdf
NERSC関係者の I/O関連基礎研究・解析:
"Parallel I/O Performance: From Events to Ensembles"
Andrew Uselton
IEEE International Parallel and Distributed Processing Symposium (IPDPS),
April 19-23, 2010
http://www.nersc.gov/projects/presentations/IPDPS2010Uselton.pptx.pdf
http://www.nersc.gov/projects/reports/technical/UseltonIPDPSpaper.pdf
"MPI-I/O on Franklin XT4 System at NERSC"
Katie Antypas and Andrew Uselton, Cray User's Group Meeting 2009
http://www.nersc.gov/projects/presentations/AntypasCUG09.pdf
"Deploying Server-side File System Monitoring at NERSC"
Andrew Uselton, Cray User's Group Meeting 2009
http://www.cug.org/7-archives/previous_conferences/CUG2009/bestpaper/16A-Uselton/uselton_slides.pdf
http://www.cug.org/5-publications/proceedings_attendee_lists/CUG09CD/S09_Proceedings/pages/authors/16-18Thursday/16A-Uselton/uselton.pdf
"MPI-I/O on Franklin XT4 System at NERSC"
Katie Antypas, Cray User's Group Meeting 2009
http://www.cug.org/5-publications/proceedings_attendee_lists/CUG09CD/S09_Proceedings/pages/authors/11-15Wednesday/13A-Antypas/Antypas-paper.pdf
"Monitoring I/O Performance on Lustre"
Andrew Uselton, Lustre User Group 2009
http://www.nersc.gov/projects/presentations/UseltonLUG2009.pdf
※こういう地道な解析が HDF5等のパフォーマンス向上に繋がってます。
※他にも関連研究はいくつもあります。
0 件のコメント:
コメントを投稿