IBM general parallel file system - introduction

IBM General Parallel File System (GPFS™) 3.5
and GSS Introduction

Karl Hansen, Nordic HPC/Technical Computing Sales Manager
IBM Systems & Technology Group
A New Era in Technical Computing: Powerful. Comprehensive. Intuitive.
IBM Confidential

Technical Computing: Powerful. Comprehensive. Intuitive

The IBM General Parallel File SystemTM (GPFSTM)
Shipping since 1998

Extreme Scalability
File system
263 files per file system
Maximum file system
size: 299 bytes
Production 19PB file
system
Number of nodes

1 to 8192

Proven Reliability

Manageability

No special nodes
Add/remove nodes
and storage on the fly
Rolling upgrades
Administer from any
node
Data replication
Snapshots

Integrated tiered storage
Storage pools
Quotas
Policy-Driven automation
Clustered NFS
SNMP monitoring
TSM / HPSS (DMAPI)

File system journaling
© 2012 IBM Corporation


IBM General Parallel File System (GPFS)

IBM General Parallel File
System (GPFS) is
a scalable highperformance file
management
infrastructure for
AIX®, Linux® and
Windows™ systems.

3

A highly available cluster architecture

Concurrent shared disk access to
a global namespace

Capabilities for high performance
parallel workloads



File data infrastructure optimization

Databases

GPFS enables:
A global namespace
across platforms
High performance
common storage
Eliminating copies
of data
Improved storage
utilization
Simplified file
management

4

Connections
SAN TCP/IP
InfiniBand
File servers

Management
Centralized
Monitoring
Automated
File Mgmt
Availability
Data Migration
Replication
Backup

Backup
and
archive

Application
servers



How is GPFS different?

Massive namespace support

GPFS
SAN

Centrally deployed, managed,
backed up and grown.

Seamless capacity and
performance scaling
All features are included. All software features: snapshots, replication and multi-site
connectivity are included in the GPFS license. With no license keys except for client and
server to add on, you get all of the features up front.

5



Network-based block input output

Application data access on network
attached nodes is exactly the same as
a SAN attached node. GPFS
transparently sends the block level IO
request over a TCP/IP network.

NSD clients

LAN

NSD
servers

GPFS
SAN

SAN

Why?
Enable virtually seamless multi-site operations
Reduce costs for data administration
Provide flexibility of file system access
Establish highly scalable and reliable data storage
Future protection by supporting mixed technologies

6



IBM General Parallel File System (GPFS™) – History & Evolution
GPFS 2.1-2.3

First
called
GPFS
HPC
GPFS
General File
Serving
Standards
Portable
operating
system
interface
(POSIX)
semantics
-Large block
Directory and
small file perf
Data
management

1998

Linux®
Clusters
(Multiple
architectures)
IBM AIX®
Loose Clusters

GPFS 3.3

GPFS 3.4

GPFS 3.5

HPC

Virtual
Tape Server
(VTS)

GPFS 3.1-3.2
Information
lifecycle
management (ILM)

Restricted
admin functions

Enhanced
Windows cluster
support
- Homogenous
Windows server

Caching via
Active File
Management
(AFM)

Research
Visualization
Digital media
Seismic
Weather
Life sciences
32 bit /64 bit
Inter-op (IBM AIX
& Linux)
GPFS Multicluster

Storage pools
File sets
Policy engine
Ease of
administration
Multiplenetworks/ RDMA
Distributed token
management

GPFS over wide
area networks
(WAN)
Large scale
clusters
thousands of
nodes

2002

Windows 2008

Improved
installation
New license
model
Improved
snapshot and
backup
Improved ILM
policy engine

GSS - GPFS
Storage Server
Performance and
scaling
improvements

GPFS File
Placement
Optimizer (FPO)

Enhanced
migration and
diagnostics
support

NFS v4 support

2005

Multiple NSD
servers

Small file
performance

2006

2009

2010

2012


A Disruptive HPC Play - GPFS Storage Server (GSS) At a Glance
The New High Capacity, High PerformanceStorage Solution

New Storage Solution fulfilled exclusively through IBM Intelligent
Cluster
High capacity, High performance, High Value offering

Product importance
Single, integrated, fully supported IBM solution
Built to leverage a strong GPFS software market
High capacity, scalable building-block approach - performance and
capacity increases as you add multiple building blocks
Cost competitive
Extreme data integrity and reduced latency with faster rebuild times

8



GPFS Storage Server - Product Description
GSS is a new storage solution fulfilled exclusively through the IBM Intelligent Cluster

Two x3650 servers combined with either four or six JBODs
Two models: GSS 24 and GSS 26
– GSS 24 (Entry): 4 JBODs – starts at nearly 500TB of storage
space
– GSS 26 (Main): 6 JBODs – starts at over 700TB of storage
space

Data
striped
across
all disks

2 and 3 TB options
10GbE or FDR Infiniband interconnects, or both!
Scalable Building Block approach to HPC Storage - performance
and capacity increase as you add multiple building blocks
Complete Storage Solution with no Storage controllers

GSS 24 Model

De-Clustered RAID Techniques
Built on GPFS software
Industry standard components
– Leverages standard components including x3650s, NetApp
JBODs, LSI SAS cards and lots of HDDs, and Intelligent
Cluster fulfillment as a single, integrated, fully supported IBM



IBM System x GPFS Storage Server provides a comprehensive
storage solution with a scalable building block approach
Storage solution includes Data Servers,
Disk (2TB or 3TB NL-SAS, SSD), Software,
InfiniBand / Ethernet with no Storage Controllers

x3650 M4 Server

JBOD
Disk Enclosure

GSS 24: Light and Fast

GSS 26: HPC Workhorse

High-Density HPC Option

2 3650 servers +
4 JBOD 20U rack
10 GB/Sec

2 3650 servers +
6 JBOD Enclosures, 28U
12 GB/sec

6 3650 servers + 18 JBOD
2 - 42U Standard Racks
36 GB/sec



Why you should care…
Affordably Scalable Building Block
approach to HPC Storage

Performance and capacity increases as you add
multiple buidling blocks
Start Small and Scale via incremental additions
Add capacity AND bandwidth

Cost competitive

Extreme data integrity and reduced
latency

Built on GPFS

Fewer parts means lower cost
Leverages System x servers and Commercial JBODs
• Fast rebuild times and industry-leading performance
• Better sustained performance
• Industry-leading throughput using efficient DeClustered RAID Techniques
• The Infrastructure for Global Technical Computing
Data Management

Single, integrated, fully supported IBM solution

Fully integrated, fully supported

Complete Storage Solution with no Storage
controllers
Easy to order through Intelligent Cluster

11



A data management portfolio for Technical Computing
Focus:
Managed Building
Block
Focus:
Ease of Use
Reliability

Focus:
Raw Raw Performance
I/O Bandwidth

IBM Data
Management
Leadership

Government

High
GPFS
Petroleum Storage Server End
Research
Media/Ent.
Financial

SONAS
Bio/Life Science
CAE

Services
Higher End
University
DCS3700
DCS3700+

Direct Attached
Smaller
(DS3500 + V3700)
Installations

IBM Tape, Tivoli Storage Manager, and HPSS



What makes this different
Clients
Clients

FDR IB
10 GbE
File/Data Servers

NSD File Server 1
x3650
NSD File Server 2

Custom Dedicated
Disk Controllers

JBOD Disk Enclosures

NSD File Server 1
GPFS Native RAID

Migrate RAID
and Disk
Management to
Standard File
Servers!

NSD File Server 2
GPFS Native RAID

JBOD Disk Enclosures



System x
Smarter Systems for a Smarter Planet

14



For more information
ibm.com/systems/software/gpfs

Email gpfs@us.ibm.com or contact your IBM Representative


IBM general parallel file system - introduction

More Related Content

What's hot

Similar to IBM general parallel file system - introduction

More from IBM Danmark

Recently uploaded

IBM general parallel file system - introduction