Information about http://www.sun.com/software/products/lustre/datasheet.pdf
LustreTM File System …
Tags: broad support, cluster file system, complex computing, data storage, gnu general public license, groundbreaking, heterogeneous networking, intensive applications, license gpl, lustre, network architecture, networking api, open networking, open source software, performance storage, petabytes, retrieval system, scalability, storage architecture, throughput,Pages: 2
Language: english
Created: Tue Jun 3 15:41:43 2008
Display cached document
Page 1
Page 2
LustreTM File System
High-performance storage architecture
and scalable cluster file system
< The LustreTM file system redefines I/O performance and scalability standards for the
world's largest and most complex computing environments. Ideally suited for data-
intensive applications that require the highest possible I/O performance, Lustre is an
object-based cluster file system that scales to tens of thousands of nodes and petabytes of storage
with groundbreaking I/O and metadata throughput.
Unique storage architecture Open source
The highly scalable and distributed Lustre file Lustre technology has been developed and
system combines open standards, the Linux maintained as open source software under the
operating system, an open networking API, and GNU General Public License (GPL), enabling
innovative protocols. Together, these elements broad support for industry-standard platforms.
create the worlds largest "network-neutral"
Heterogeneous networking
data storage and retrieval system.
The Lustre network architecture provides
Applying intelligence throughout its architec- flexible support for a wide variety of networks
ture, the Lustre file system turns commodity and high-performance features. The Lustre file
Highlights hardware into smart storage devices that system interoperates with network vendor-
· Unparalleled scalability: Employs manage data objects. The objects are dynami- supplied libraries through Lustre Network
an object-based storage architecture cally distributed horizontally across the servers, Drivers, which utilize advanced features such
that scales to tens of thousands of shattering the performance limitations of as Remote Direct Memory Access, OS-bypass
clients and petabytes of data--a file traditional storage systems. for parallel I/O, and vector I/O for efficient
system virtually without limits bulk data movement. Lustre Network Drivers
Building a Lustre cluster requires a Lustre
· Reliable: Deployed in production exist for many networks, including TCP/IP,
on many large and small clusters, MetaData Server and Lustre Object Storage
Quadrics Elan, many flavors of InfiniBand, and
meeting the uptime requirements Servers, each with disk storage. A pool of client
Myrinet GM--each with performance excep-
of business and national security systems can access these servers through one
applications tionally close to the raw device throughput.
of many supported networks. The Lustre file
· Proven performance: Delivers operations bypass the MetaData Server com- Rugged high availability
dramatic increases in throughput and pletely and fully utilize the parallel data paths The Lustre file system organizes all servers in
I/O by intelligent serialization and to all Object Storage Servers in the cluster. This active-active failover pairs. Together with
separation of metadata operations from
unique approach of separating metadata protocol interoperability between versions,
data manipulation
operations from data operations results in live cluster upgrades are now routine.
· Cost effective: Significantly reduces
significantly enhanced performance.
deployment and support costs through POSIX compliance
support for industry-standard platforms Like other UNIX® and Linux file systems, The Lustre file system provides a tested and
and heterogeneous networking
Lustre files are represented by inodes. fully compliant file system interface in
environments
However, a key difference in the Lustre file accordance with the POSIX standard.
· Open source, open standards: Developed system is that its inodes simply contain refer-
and maintained as open source software
ences to the objects storing the file data.
with an open networking protocol and
POSIX file system semantics, ensuring
broad support for industry- standard
platforms and heterogeneous
networking environments
2 Datasheet LustreTM File System sun.com/lustre
Simple configuration Lustre File System Specifications
With version 1.6, the Lustre file system The Lustre file system version 1.6 is now Learn More
revolutionizes configuration simplicity. Routine available to the general public under the GNU GPL.
For more information on the
formatting and mounting of server devices
Operating systems Lustre File System, visit
aggregates them into a global high-availability
· Red Hat Enterprise Linux 3 and later sun.com/lustre or www.lustre.org
cluster file system.
· SUSE Linux Enterprise Server 9 and 10
Innovative protocols
· Linux 2.4 and 2.6
The Lustre file system employs a distributed
· SolarisTM Operating System (coming in 2008)
lock manager to handle access to files and
directories and synchronize updates. This SunSpectrum support and services
Hardware platforms
feature improves the metadata journaling · Intel Architecture 32-bit/64-bit Sun provides comprehensive software
approach used by most modern file systems. support for your Lustre application with the
· x86/x64
SunSpectrum Software Service Plan at the
· PowerPC
Extreme parallel computing Standard level. Having access to Sun's expert
The Lustre lock manager automatically adapts Networking technicians, in-depth knowledge base, software
its policies to minimize overhead for the current · Quadrics Elan-3 and Elan-4 updates and releases helps ensure the success
application. Files being used by a single node of your deployment and the availability of your
· TCP/IP
are covered by a single lock, eliminating solution.
· InfiniBand
additional lock overhead. Nodes sharing files
get the largest possible locks, which still allows · Myrinet For more information see:
all nodes to write at full speed. sun.com/service/serviceplans/software
Proven performance
Intent-based locking Lustre technology powers most of the world's And, to further ensure the success of your
largest Linux supercomputers and is the first project, take advantage of Lustre administra-
To dramatically reduce bottlenecks and to production-tested, object-based Linux cluster file
increase overall data throughput, the Lustre file system. tion training. In a packed 3-day course, Sun
system uses an intent-based locking mechanism, Learning Services covers best practices in
Recent results
where file and directory lock requests also deployment and administration as well as
· File I/O % of raw bandwidth: >90%
provide information about the reason the lock configurations from very simple to very
· Achieved single OSS I/O: >2.5 GB/s complex.
is being requested. For example, if a directory
lock is being requested to create a new unique · Achieved single client I/O: >2.0 GB/s
Featured partners
file, the Lustre file system handles this as a · Single GigE end-to-end throughput:
· Bull · Quadrics
single request. In other file systems, this action 118 MB/s
· Cray · Scali
requires multiple network requests for lookup, · Achieved aggregate I/O: 130 GB/s
· DataDirect · Dell
creation, opening, and locking. · Metadata transaction rate: 15,000 ops/s
Networks · Red Hat
· Maximum clients supported: 25,000 · HP · LSI
· Maximum file/file system size: · Hitachi Data · EMC
1.25 PB/>32 PB Systems · SGI
· Linux Networx · SiCortex
· Novell · Terascala
Sun Microsystems, Inc. 4150 Network Circle, Santa Clara, CA 95054 USA Phone 1-650-960-1300 or 1-800-555-9SUN Web sun.com
© 2008 Sun Microsystems, Inc. All rights reserved. Sun, Sun Microsystems, the Sun logo, Lustre, the Lustre logo, and Solaris are trademarks or registered trademarks of Sun Microsystems, Inc. or its subsidiaries in the United
States and other countries. Information subject to change without notice. SunWIN #518547 Lit. #SYDS13587-1 06/08