DDN s Vision for the Future of Lustre LUG2015 Robert Triendl

Similar documents
Leveraging Software-Defined Storage to Meet Today and Tomorrow s Infrastructure Demands

SFA12KX and Lustre Update

HPC Storage Use Cases & Future Trends

DDN. DDN Updates. DataDirect Neworks Japan, Inc Nobu Hashizume. DDN Storage 2018 DDN Storage 1

DDN and Flash GRIDScaler, Flashscale Infinite Memory Engine

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

DDN. DDN Updates. Data DirectNeworks Japan, Inc Shuichi Ihara. DDN Storage 2017 DDN Storage

朱义普. Resolving High Performance Computing and Big Data Application Bottlenecks with Application-Defined Flash Acceleration. Director, North Asia, HPC

DDN About Us Solving Large Enterprise and Web Scale Challenges

LCE: Lustre at CEA. Stéphane Thiell CEA/DAM

Enhancing Lustre Performance and Usability

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

2012 HPC Advisory Council

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

Next-Generation NVMe-Native Parallel Filesystem for Accelerating HPC Workloads

Lustre2.5 Performance Evaluation: Performance Improvements with Large I/O Patches, Metadata Improvements, and Metadata Scaling with DNE

Emerging Technologies for HPC Storage

Lustre / ZFS at Indiana University

Data Movement & Tiering with DMF 7

CSCS HPC storage. Hussein N. Harake

Lustre usages and experiences

I/O at the Center for Information Services and High Performance Computing

Lustre overview and roadmap to Exascale computing

Andreas Dilger. Principal Lustre Engineer. High Performance Data Division

Active-Active LNET Bonding Using Multiple LNETs and Infiniband partitions

IME (Infinite Memory Engine) Extreme Application Acceleration & Highly Efficient I/O Provisioning

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

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

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

SGI Overview. HPC User Forum Dearborn, Michigan September 17 th, 2012

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

Xyratex ClusterStor6000 & OneStor

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

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

Application Performance on IME

Campaign Storage. Peter Braam Co-founder & CEO Campaign Storage

Quobyte The Data Center File System QUOBYTE INC.

IME Infinite Memory Engine Technical Overview

Extraordinary HPC file system solutions at KIT

NetApp: Solving I/O Challenges. Jeff Baxter February 2013

Robin Hood 2.5 on Lustre 2.5 with DNE

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

Data Movement & Storage Using the Data Capacitor Filesystem

NetApp High-Performance Storage Solution for Lustre

Data Analytics and Storage System (DASS) Mixing POSIX and Hadoop Architectures. 13 November 2016

Lustre Metadata Fundamental Benchmark and Performance

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

Data Management. Parallel Filesystems. Dr David Henty HPC Training and Support

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

Using DDN IME for Harmonie

Andreas Dilger, Intel High Performance Data Division LAD 2017

Isilon: Raising The Bar On Performance & Archive Use Cases. John Har Solutions Product Manager Unstructured Data Storage Team

High Capacity network storage solutions

Lustre* is designed to achieve the maximum performance and scalability for POSIX applications that need outstanding streamed I/O.

Accelerating Spectrum Scale with a Intelligent IO Manager

Feedback on BeeGFS. A Parallel File System for High Performance Computing

A ClusterStor update. Torben Kling Petersen, PhD. Principal Architect, HPC

Improved Solutions for I/O Provisioning and Application Acceleration

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

MAHA. - Supercomputing System for Bioinformatics

Fujitsu s Contribution to the Lustre Community

HIGH-PERFORMANCE STORAGE FOR DISCOVERY THAT SOARS

Andreas Dilger, Intel High Performance Data Division SC 2017

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

INTEGRATING HPFS IN A CLOUD COMPUTING ENVIRONMENT

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

Challenges in making Lustre systems reliable

ZEST Snapshot Service. A Highly Parallel Production File System by the PSC Advanced Systems Group Pittsburgh Supercomputing Center 1

Refining and redefining HPC storage

Analyzing I/O Performance on a NEXTGenIO Class System

Jessica Popp Director of Engineering, High Performance Data Division

Lustre architecture for Riccardo Veraldi for the LCLS IT Team

Community Release Update

Architecting Storage for Semiconductor Design: Manufacturing Preparation

Isilon Performance. Name

libhio: Optimizing IO on Cray XC Systems With DataWarp

The modules covered in this course are:

IBM Spectrum Scale vs EMC Isilon for IBM Spectrum Protect Workloads

High Performance Computing. NEC LxFS Storage Appliance

Andreas Dilger High Performance Data Division RUG 2016, Paris

Architecting a High Performance Storage System

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

Brent Gorda. General Manager, High Performance Data Division

ENERGY-EFFICIENT VISUALIZATION PIPELINES A CASE STUDY IN CLIMATE SIMULATION

Parallel File Systems. John White Lawrence Berkeley National Lab

Introduction to Lustre* Architecture

An Exploration into Object Storage for Exascale Supercomputers. Raghu Chandrasekar

Fast Forward I/O & Storage

Got Isilon? Need IOPS? Get Avere.

Lustre HSM at Cambridge. Early user experience using Intel Lemur HSM agent

Sun Lustre Storage System Simplifying and Accelerating Lustre Deployments

Storage Evaluations at BNL

InfiniBand Networked Flash Storage

The Btrfs Filesystem. Chris Mason

DVS, GPFS and External Lustre at NERSC How It s Working on Hopper. Tina Butler, Rei Chi Lee, Gregory Butler 05/25/11 CUG 2011

Computer Science Section. Computational and Information Systems Laboratory National Center for Atmospheric Research

Sharing High-Performance Devices Across Multiple Virtual Machines

DELL EMC ISILON F800 AND H600 I/O PERFORMANCE

All-Flash High-Performance SAN/NAS Solutions for Virtualization & OLTP

System Software for Big Data and Post Petascale Computing

Transcription:

DDN s Vision for the Future of Lustre LUG2015 Robert Triendl

3 Topics 1. The Changing Markets for Lustre 2. A Vision for Lustre that isn t Exascale 3. Building Lustre for the Future 4. Peak vs. Operational Performance 5. Application Optimized Lustre 6. Why Conventional Storage Still Matters

4 Hyperscale Storage Markets HPC Scratch Petabytes Streaming Write Large Files Infiniband Big Data & Data Analytics Cloud WORM(N) Billions of Files Random Read Small Files Ethernet Single Location Distributed

5 Lustre Markets Today Cloud 2% Archive 12% Work 31% Mixed 22% Data 33%

6 Market Diversification Cloud Work Data Mixed Use Archive Weather Climate HPC Work CAE Chemical General Academic HPC Cloud Tier 2 Genomics Finance Cloud Big Data Science Energy Security

7 Lustre Futures Beyond Exascale Manufacturing Genomics General Academic Cloud Archive CIFS/NFS Export, AD Integration, RAS Features, Snapshots, Data Management, etc. Random Performance, Small File & Metadata Performance, Data Management, Security, etc. Broad Application Support, Connectors, User Monitoring, User Access to Snapshot, etc. Virtualization, Snapshots, Small File Read Performance, Data Distribution, etc. Data Management Features, SMR Drive Use, Data Scrubs, Data Distribution, etc.

8 Market Evolution 100% 90% 80% 70% 60% 50% 40% 30% 20% Archive Cloud Mixed Data Work 10% 0% 2012 2013 2014

9 Market Segments 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 2011 2012 2013 2014 Industry Government University

10 Disks: Throughput vs. IOPS MB/sec 1800 1600 1400 1200 1000 800 600 400 200 IOPS/10 GB/sec 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0 Lustre 1.8 Lustre 2.4 ExaScaler 2.2 Lustre with BtrFS Next Gen SAS Drives 0

11 Lustre Development at DDN Lustre Usability Features Build-in Reliability and Availability Lustre Recovery Features for a Broader Market Performance for Broad Set of Applications Application-optimized Lustre

12 Lustre Code Contributions 120 100 80 60 40 Other EMC SUSE Bull Cray Seagate DDN 20 0 2.1 2.4.0 2.3.50-2.4.0 2.5.0 2.5.50-2.6.0

13 DDN ExaScaler Software Stack DDN DDN & Intel Intel HPDD Other Data Management Fast Data Copy Tape S3 Cloud Object (WOS) ExaScaler Data Management Framework Monitoring & Management DDN DirectMon DDN ExaScaler DDN ExaScaler Monitor Intel IML DDN Clients NFS/CIFS/S3 DDN IME Intel Hadoop Core FS ldiskfs Intel DSS DDN Lustre Edition OpenZFS btrfs Storage HW DDN Block Storage with SFX Cache Other HW

14 Why BtrFS? Standard Local Filesystem in RHEL7 Better Throughput Performance than ZFS Similar Feature Set, but all Linux No Possible Patent Infringement Simple Integration and Deployment

15 Application-Optimized Lustre Lustre for Specific Applications Workload Profiling Optimization Across I/O Calls Optimizing Application Runtime Working with Customers

16 Genome Pipeline Benchmarks Samtools 20% faster with DDN Lustre optimizations Runtime (Hours) Lustre 2.5 Client Performance Human Genetics samtools workflow 50 40 30 20 10 0 2.5.1 DDN Branch

17 SSD Pools and Caching DSS to Link File Layer to Block Layer Build into the File System Better use of SSDs for I/O Optimization Increased Small File Performance Increased Random Read to Large Files Additional specificity with fadvice()

18 ExaScaler Monitoring Filesystem, OSS, MDS, OST, MDT, etc. JOB ID, UID/GID, application stats, etc. Archive of data by policy Lightweight Near real- Rme Massive scale Burst Buffer Monitoring Server collectd Graphite plugin OSS, MDS Storage UDP(TCP)/IP based small text message transfer graphite

19 TITECH Examples 64 Billion of Lustre Stats in 15 days!

20 Why Block-Level Raid? Best Mixed I/O Performance Consistent Performance Hardware-optimized Performance Best Performance During Failure Integrated Storage Services

21 SFA RAID Stack Performance Above 1 Million 4K IOPS per 8 CPU Cores Above 10 GB/sec per 8 CPU Cores 8 Cores Sufficient for PCI Infrastructure More Cores for File System Services Additional Cores for More Functionality

22 SFA Random Read MB/sec 30 25 20 15 10 5 512K I/O Size 1M I/O Size 2M I/O Size 4M I/O Size 0 1 8 16 24 32 40 48 56

23 Flexible SSU Design S M L XL

24 SFA14K Performance SSU Up to 45 GB/sec Up to 2950 TB External MDT 4-6 OSS Monitoring 2-4 MDS MDT Storage OST Storage

25 Wolfcreek Hardware

26 Wolfcreek Hardware