Community Release Update

Similar documents
Community Release Update

Jessica Popp Director of Engineering, High Performance Data Division

Andreas Dilger. Principal Lustre Engineer. High Performance Data Division

Andreas Dilger, Intel High Performance Data Division SC 2017

Lustre * Features In Development Fan Yong High Performance Data Division, Intel CLUG

Andreas Dilger, Intel High Performance Data Division LAD 2017

Andreas Dilger, Intel High Performance Data Division Lustre User Group 2017

Open SFS Roadmap. Presented by David Dillow TWG Co-Chair

DL-SNAP and Fujitsu's Lustre Contributions

Improved Versioning, Building, and Distribution of Lustre

Fujitsu s Contribution to the Lustre Community

Parallel File Systems Compared

Lustre * 2.12 and Beyond Andreas Dilger, Intel High Performance Data Division LUG 2018

GPFS on a Cray XT. Shane Canon Data Systems Group Leader Lawrence Berkeley National Laboratory CUG 2009 Atlanta, GA May 4, 2009

LustreFS and its ongoing Evolution for High Performance Computing and Data Analysis Solutions

Lustre 2.12 and Beyond. Andreas Dilger, Whamcloud

Lustre overview and roadmap to Exascale computing

Multi-tenancy: a real-life implementation

Demonstration Milestone Completion for the LFSCK 2 Subproject 3.2 on the Lustre* File System FSCK Project of the SFS-DEV-001 contract.

<Insert Picture Here> Lustre Community

Upstreaming of Features from FEFS

6.5 Collective Open/Close & Epoch Distribution Demonstration

Small File I/O Performance in Lustre. Mikhail Pershin, Joe Gmitter Intel HPDD April 2018

Olaf Weber Senior Software Engineer SGI Storage Software. Amir Shehata Lustre Network Engineer Intel High Performance Data Division

LUG 2012 From Lustre 2.1 to Lustre HSM IFERC (Rokkasho, Japan)

Introduction The Project Lustre Architecture Performance Conclusion References. Lustre. Paul Bienkowski

Scalability Testing of DNE2 in Lustre 2.7 and Metadata Performance using Virtual Machines Tom Crowe, Nathan Lavender, Stephen Simms

Lustre / ZFS at Indiana University

Intel Enterprise Edition Lustre (IEEL-2.3) [DNE-1 enabled] on Dell MD Storage

Shared Object-Based Storage and the HPC Data Center

DDN s Vision for the Future of Lustre LUG2015 Robert Triendl

Multi-Rail LNet for Lustre

LUSTRE NETWORKING High-Performance Features and Flexible Support for a Wide Array of Networks White Paper November Abstract

Olaf Weber Senior Software Engineer SGI Storage Software

High-Performance Lustre with Maximum Data Assurance

The modules covered in this course are:

The current status of the adoption of ZFS* as backend file system for Lustre*: an early evaluation

SFA12KX and Lustre Update

HPSS Development Update and Roadmap 2017

Intel Omni-Path Fabric Manager GUI Software

Demonstration Milestone for Parallel Directory Operations

LNet Roadmap & Development. Amir Shehata Lustre * Network Engineer Intel High Performance Data Division

LNET MULTI-RAIL RESILIENCY

Johann Lombardi High Performance Data Division

Lustre Metadata Fundamental Benchmark and Performance

Fujitsu's Lustre Contributions - Policy and Roadmap-

Parallel File Systems. John White Lawrence Berkeley National Lab

UK LUG 10 th July Lustre at Exascale. Eric Barton. CTO Whamcloud, Inc Whamcloud, Inc.

Administering Lustre 2.0 at CEA

Introduction to Lustre* Architecture

Oak Ridge National Laboratory Computing and Computational Sciences

LLNL Lustre Centre of Excellence

Lustre Interface Bonding

Lustre* - Fast Forward to Exascale High Performance Data Division. Eric Barton 18th April, 2013

Toward a Windows Native Client (WNC) Meghan McClelland LAD2013

HPSS Treefrog Summary MARCH 1, 2018

Lustre A Platform for Intelligent Scale-Out Storage

Enhancing Lustre Performance and Usability

Analytics of Wide-Area Lustre Throughput Using LNet Routers

EMC Lustre Contributions

Lustre at the OLCF: Experiences and Path Forward. Galen M. Shipman Group Leader Technology Integration

Lustre usages and experiences

A GPFS Primer October 2005

Architecting a High Performance Storage System

DPDK Roadmap. Tim O Driscoll & Chris Wright Open Networking Summit 2017

Coordinating Parallel HSM in Object-based Cluster Filesystems

Extraordinary HPC file system solutions at KIT

Implementing Storage in Intel Omni-Path Architecture Fabrics

Lustre on ZFS. Andreas Dilger Software Architect High Performance Data Division September, Lustre Admin & Developer Workshop, Paris, 2012

<Insert Picture Here> Lustre Development

Arm in HPC. Toshinori Kujiraoka Sales Manager, APAC HPC Tools Arm Arm Limited

Lustre at Scale The LLNL Way

Experiences with HP SFS / Lustre in HPC Production

Lustre on ZFS. At The University of Wisconsin Space Science and Engineering Center. Scott Nolin September 17, 2013

Mission-Critical Lustre at Santos. Adam Fox, Lustre User Group 2016

An Overview of Fujitsu s Lustre Based File System

Filesystems on SSCK's HP XC6000

Simplified Multi-Tenancy for Data Driven Personalized Health Research

Parallel File Systems for HPC

Fan Yong; Zhang Jinghai. High Performance Data Division

Improving overall Robinhood performance for use on large-scale deployments Colin Faber

Lustre File System. Proseminar 2013 Ein-/Ausgabe - Stand der Wissenschaft Universität Hamburg. Paul Bienkowski Author. Michael Kuhn Supervisor

InfiniBand Strengthens Leadership as the Interconnect Of Choice By Providing Best Return on Investment. TOP500 Supercomputers, June 2014

RobinHood Project Update

Roads to Zester. Ken Rawlings Shawn Slavin Tom Crowe. High Performance File Systems Indiana University

Store Process Analyze Collaborate Archive Cloud The HPC Storage Leader Invent Discover Compete

Birds of a Feather Presentation

A Comparative Experimental Study of Parallel File Systems for Large-Scale Data Processing

vnetwork Future Direction Howie Xu, VMware R&D November 4, 2008

Compute Node Linux: Overview, Progress to Date & Roadmap

Oak Ridge National Laboratory

The Last Bottleneck: How Parallel I/O can improve application performance

Digitizer operating system support

A Generic Methodology of Analyzing Performance Bottlenecks of HPC Storage Systems. Zhiqi Tao, Sr. System Engineer Lugano, March

FhGFS - Performance at the maximum

Managing HPC Active Archive Storage with HPSS RAIT at Oak Ridge National Laboratory

An Introduction to GPFS

Open Enterprise & Open Community opensuse & SLE Empowering Each Other. Richard Brown opensuse Chairman

Illinois Proposal Considerations Greg Bauer

Data life cycle monitoring using RoBinHood at scale. Gabriele Paciucci Solution Architect Bruno Faccini Senior Support Engineer September LAD

Transcription:

Community Release Update LUG 2017 Peter Jones HPDD, Intel OpenSFS Lustre Working Group

OpenSFS Lustre Working Group Lead by Peter Jones (Intel) and Dustin Leverman (ORNL) Single forum for all Lustre development matters Oversees entire Lustre development cycle Maintains the roadmap Plans major releases Collects requirements for future Lustre features Sets priorities for test matrix For more information visit the wiki http://wiki.opensfs.org/lustre_working_group 2

Community Survey Has been running for 6 years LWG devises questions Useful for tracking trends in Lustre usage Drop in respondents compared to 2016 survey (97 to 70) Full details available at http://wiki.opensfs.org/lustre_co mmunity_survey 120 100 80 60 40 20 0 Respondents 2012 2013 2014 2015 2016 2017 3

Community Survey - Versions 70.00% Which Lustre Version(s) Are You Running in Production? 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% Lustre 1.8.x Lustre 2.1.x Lustre 2.4.x Lustre 2.5.x Lustre 2.6 Lustre 2.7.x Lustre 2.8.x Lustre 2.9.x 2016 2017 Lustre 2.5.x remains the most commonly-used version in production Strong adoption of more current releases 4

Community Survey - Linux distros Ubuntu 16.04 Ubuntu 14.04 SLES 12 SP2 What Linux distributions do you use in production? SLES 12 SP1 SLES 12 SLES 11 SPx Clients # Servers # RHEL/CentOS 7.x RHEL/CentOS 6.x RHEL/CentOS 5.x 0 10 20 30 40 50 60 70 RHEL/CentOS 6.x still the most widely used distro RHEL/CentOS 7.x now widely used 5

Community Survey - Networks Which kind of networks do you use with Lustre within your organization? 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% 1 GigE 10 GigE 40 GigE Linux distribution IB Mellanox IB OFA IB Intel True Scale (QLogic) IB Intel Omnipath (OPA) Many options used but Mellanox and 10 GigE remain the most common 6

Community Survey New Features ZFS Wide Striping (161-2000 stripes) What describes your level of interest in...? SSK LFSCK (lctl) Will use within year Use in production NRS HSM DNE 0 5 10 15 20 25 30 DNE and LFSCK usage grew since last year 7

Community Survey - Support What best describes your Lustre support situation? We have formal Lustre support arrangements in place with another organization We self-support We offer support to others Even split between those self-supporting and those using/providing vendor support 8

Lustre 2.9 GA December 2016 Supports RHEL 7.3 servers/clients and SLES12 SP1 clients Varied features UID/GID mapping (LU-3291) Shared Secret Key Crypto (LU-3289) Subdirectory Mounts (LU-28) ZFS Metadata Improvements (LU-7895) RPM Kernel Module Weak Updates (LU-5614) Server Side I/O Advise (ladvise) (LU-4931) Large Bulk IO (LU-7990) http://wiki.lustre.org/release_2.9.0 9

Lustre 2.9 - Contributions Number of Commits Lines of Code Changed Purdue 1 Other 11 SGI 1 ORNL 43 LLNL 34 ANU 14 Atos 7 CEA 8 Clogeny 1 Seagate 93 Cray 36 IU 36 DDN 24 Fujitsu 1 Purdue 43 SGI 305 Other 879 ORNL 7873 LLNL 4184 ANU 545 Seagate 5138 Atos 449 Cray 3671 CEA 210 DDN 2857 IU 9319 Clogeny 15 Fujitsu 1 Intel 427 Intel 38075 Statistics courtesy of Dustin Leverman (ORNL) Aggregated data by organization between 2.8.50 and 2.9.0 tags Source: http://git.whamcloud.com/fs/lustre-release.git/shortlog/refs/heads/b2_9 10

Lustre 2.9 - Reviews SGI 8 Other 2 LLNL 17 Number of Reviews ANU 4 Atos 6 CEA17 Cray Seagate 75 89 DDN 53 GSI 1 ORNL 142 IU 4 Intel 1821 Source: Statistics courtesy of Dustin Leverman (ORNL) Aggregated data by organization between 2.8.50 and 2.9.0 tags 11

Lustre 2.10 Targeted GA June 2017 Latest status on how this is trending Will support RHEL 7.3 servers/clients and SLES12 SP2 clients Interop/upgrades from Lustre 2.9 servers/clients Will be designated an LTS Release and have freely available maintenance releases Lustre 2.10.1 targeted for Q3 release Lustre 2.10.x will remain LTS branch for at least 18 months http://wiki.lustre.org/release_2.10.0 12

Lustre 2.10 - Features A number of new features included in this release Progressive File Layouts (LU-8998) Multi-Rail LNET (LU-7734) Project Quotas (LU-4017) NRS Delay Policy (LU-6283) ZFS Snapshots (LU-8900) http://wiki.lustre.org/release_2.10.0 13

Lustre 2.10 Progressive File Layouts Progressive File Layout (PFL) simplifies usage for users and admins Optimize performance for diverse users/applications One PFL layout could be used for all files Low stat overhead for small files High IO bandwidth for large files Collaboration between Intel and ORNL Example progressive file layout with 3 components 1 stripe [0, 32MB) 4 stripes [32MB, 1GB) 128 stripes [1GB, ) 14

Lustre 2.10 Multi-Rail LNet Allow LNet across multiple network interfaces Supports all LNet networks LNet layer instead of LND layer Allows concurrent use of different LNDs (e.g. both TCP & IB at once) Scales performance significantly Improves reliability Active-active network links between peers Collaboration between Intel and HPE/SGI 15

Lustre 2.10.x Additional Content Confirmed in 2.10.0 ZFS Metadata Improvements (LU-7895) OPA Performance improvements (LU-8943) Pacemaker scripts (LU-8457/8458) Feature/bugfix parity with latest EE 3.x Lustre releases Coming in 2.10.x maintenance release Patchless servers (LU-20) Support for 4.9 kernel Lustre clients (LU-9183) Single thread performance improvements (LU-8964) SLES12 SP2 server support Ubuntu 16.04 LTS Lustre client support MOFED 4.x support 16

Lustre 2.10 - Contributions Number of Commits Lines of Code Changed Hamburg 6 Atos 2 CEA 6 Cray 13 Seagate 131 Hamburg Atos 29 CEA 1184 Cray HPE/SGI 9 HPE/SGI 142 2407 GSI 1 DDN 4275 Other 22 IU 6 Seagate DDN 346 LLNL 224 ORNL GSI 23 38 30 6737 IU 142 LLNL 5 ORNL 46 Intel 418 Intel 75308 Statistics courtesy of Dustin Leverman (ORNL) Source: http://git.whamcloud.com/fs/lustre-release.git/shortlog/refs/heads/master Aggregated data by organization between 2.9.50 and 2.9.56 tags 17

Lustre 2.10 - Reviews Other 8 ORNL 72 LLNL 7 Hamburg 9 Stanford 2 HPE/SGI 26 Seagate 47 CEA 26 Cray 33 DDN 32 IU 9 Atos 2 Number of Reviews Intel 1445 Source: http://git.whamcloud.com/fs/lustre-release.git/shortlog/refs/heads/master Aggregated data by organization between 2.9.50 and 2.9.58 tags Statistics courtesy of Dustin Leverman (ORNL) 18

Lustre Version Statistics Version Commits LOC Developers Organizations 1.8.0 997 291K 41 1 2.1.0 752 92K 55 7 2.2.0 329 58K 42 10 2.3.0 586 87K 52 13 2.4.0 1123 348K 69 19 2.5.0 471 102K 70 15 2.6.0 885 147K 76 14 2.7.0 742 201K 65 15 2.8.0 995 147K 92 17 2.9.0 737 74K 121 16 2.10.0* 605 93K 81 13 Statistics courtesy of Chris Morrone (LLNL)/ Dustin Leverman (ORNL) *Data for 2.10 is incomplete and only runs up until 2.9.58 Source: http://git.whamcloud.com/fs/lustre-release.git 19

Lustre 2.11 and Beyond Many interesting long term features in development File Level Redundancy Data on MDT LNet Network Health Full list maintained on lustre.org wiki http://wiki.lustre.org/projects Roadmap kept up to date on lustre.org http://lustre.org/roadmap/ 20

Community Release Roadmap 2.9 2.10* 2.11 2.12 UID/GID Mapping Shared Key Crypto Large Block IO Subdirectory Mounts ZFS Snapshots Multi-rail LNET Progressive File Layouts Project Quotas` Data on MDT FLR Delayed Resync Lock Ahead FLR Immediate Resync LNet Network Health Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 2017 *LTS Release with maintenance releases provided Estimates are not commitments and are provided for informational purposes only Fuller details of features in development are available at http://wiki.lustre.org/projects Last updated: April 20 th 2017 2018 21

LTS Releases Long Term Stable (LTS) releases will remain active for at least 18 months Similar idea to Whamcloud maintenance release streams Updates for bugfixes and to support newer Linux distro releases Will announce details of next release as they become available Maintenance releases may also be produced for other community releases but only until next major release i.e. no updates for 2.n when 2.n+1 has been released Will depend upon levels of adoption 22

Lustre Release Documentation Latest version of user documentation dynamically available to download http://lustre.org/documentation/ Also links to details on how to contribute If you know of gaps then please open an LUDOC ticket If you have not got time to work out the correct format to submit then unformatted text will provide a starting point for someone else to complete More dynamic content on http://wiki.lustre.org 23

Summary Lustre 2.10.0 is expected shortly Freely available 2.10.x maintenance releases planned Plenty of options for those interest in contributing to Lustre 24

Thank you Open Scalable File Systems, Inc. 3855 SW 153rd Drive Beaverton, OR 97006 Ph: 503-619-0561 Fax: 503-644-6708 admin@opensfs.org www.opensfs.org