HPE Clustered Extents File System: технические характеристики и документация

HPE Clustered Extents File System 

The HPE Clustered Extents File System (HPE XFS) is a shared file system designed to provide simultaneous, high speed access to data among clustered Linux servers connected to a SAN, where each server in the cluster has direct high-speed data channels to a shared set of disks. 

The HPE XFS clients share a single namespace within the cluster, so each client can see all directoris and files, and can access them at local to near-local speeds. HPE Clustered Extents File System can scale for bandwidth or I/O by adding additional storage or network connections. It delivers High Availability (HA) of data by detecting and automatically recovering from a server or network failure. 

Key Benefits

High Performance File Sharing for clustered Linux systems.

• With the HPE Clustered Extents File System, no network mounts or file copying is involved. Multiple servers have direct access to the same file at the same time and at near local speeds.
• Streamlined metadata management is designed for performance with sophisticated algorithms and structures for fast buffering and lookups.
• Dedicated metadata servers can be utilized for increased performance. 

Highly Scalable for Capacity or Bandwidth

• The HPE Clustered Extents File System is a 64-bit file system, is able to handle individual files up to 9 Exabytes, and file system sizes up to 18 Exabytes.
• Scales bandwidth by adding additional network connections for increased flexibility and increases storage capacity by seamlessly adding additional drives.
• Dynamic allocation algorithms optimize disk resource utilization.
• Disk volumes can be allocated across thousands of disks to allow for future storage needs.
• Multiple Allocation Groups provide independence of metadata structires between files and directories placed in different groups. 

Designed for All Flash Storage

• HPE Clustered Extents File System employs I/O model with direct access for clients that dramatically reduces software layer latenciess when accessing high-speed, NVMe-based storage.
• A write-in-place design for both data and metadata avoids write amplification.
• I/Os as small as a single sector can be issued without penalty.
• HPE Clustered Extents File System clients can benefit from near-hardware IOPS performance. 

High Availability

• The HPE Clustered Extents File System detects and automatically recovers from a single system failure, including metadata server failure, as well as a network failure.
• HPE Clustered Extents File System is a journaling file system, which logs changes not yet committed to the file system, providing for file system consistency and for rapid recovery in the event of an improper system shutdown.
• Multiple backup metadata servers can be designated for failover in case of a metadata server failure.

Clustered data sharing

HPE XFS software architecture allows simultaneous shared access to storage-the systems have direct access to data and are not bottlenecked by slow, congested networks, or overloaded file servers. HPE XFS provides local-drive speed data sharing combining 64-bit scalability and performance with SAN technology, enabling computer systems to share data efficiently and do more work in less time. This allows customers to realize the benefits of SAN consolidation with a significant growth in overall productivity that increases Return on Investment (ROI).

Metadata management

Streamlined metadata management is the key to HPE XFS performance. Metadata includes information about files (size, ownership, timestamps, location, and more), as well as information about the file system itself. While all file data in HPE XFS flow directly between clients and disk, transactions that alter metadata are brokered by a metadata server for each HPE XFS instance, coordinating access and facilitating data integrity. 

Metadata transactions are routed over a TCP/IP network to the metadata server. Since metadata transactions are typically small and infrequent relative to data transactions, a Gigabit Ethernet connection is typically adequate. Faster connections such as 10-Gigabit Ethernet or InfiniBand are also supported and are recommended for high-performance environments. 

Several underlying HPE XFS features enhance metadata performance:

• Fast metadata algorithms with excellent buffering
• Sophisticated structures and algorithms for fast lookups
• Ability to allocate large extents, helping minimize metadata

A unique remote procedure call (RPC) design enables high-speed processing of metadata transactions. HPE XFS utilizes:

• Specialized thin RPCs
• Excellent buffering of both data and metadata on clients
• Multiple metadata servers (one per file system)
• Ability to bypass the HPE XFS layers on the metadata server
• Ability to dedicate small configuration servers as metadata servers 

HPE XFS uses a fast and efficient token mechanism to control file access, which enables transaction accuracy that is far better than with typical network file systems. Because file locking is critical for data integrity, HPE XFS takes great care in the implementation of POSIX®, Berkeley Software Distribution (BSD), and UNIX® System V (SVR4) file locks. No application changes are required to use file locking with HPE XFS.

High-availability

HPE XFS is designed to detect and automatically recover from single system failure, failure of the active metadata server, failure of SAN or TCP/IP network components, and failure of the network or SAN partition. In an HPE XFS configuration, one system is the metadata server and other server-capable nodes may be designated as backup metadata servers. If the initial metadata server fails, the backup metadata server takes over in seconds, which facilitates continual access to data regardless of the state of any single system

Data Management Framework Integration

HPE XFS is tightly integrated with HPE Data Management Framework (HPE DMF) version 6 and 7. It supports native DMAPI that facilitates data migrations and recalls. DMF7 leverages HPE XFS DMAPI capability to build namespace reflection that can be quireied based on metadata, including extended attribites. In addition, DMF provides filesystem backup capability. DMF movers simply mount the HPE XFS and migrate or recall data in parallel.

Complete SAN/NAS gateway

Systems not participating in the HPE XFS membership can still have access to all data. Non-participating systems simply remain in their current configurations and access data in the HPE XFS volumes through a member system via traditional data sharing methods such as SMB, NFS or FTP.

HPE Clustered Extents File System Ordering Rules

HPE XFS Base SKU

HPE Data Management Framework Suite Base 1yr Subscription E-Media

R1C83AAE

HPE XFS Shared File System (CXFS) Base SKUs

HPE Extents File System 1-server Base System 1yr Subscription E-LTU

R1D38AAE

HPE Extents File System 1-server Base System 3yr Subscription E-LTU

R1D39AAE

HPE Extents File System 1-server Base System 5yr Subscription E-LTU

R1D40AAE

HPE XFS Shared File System (CXFS) Increment SKUs

HPE Data Management Framework/Extents File System Additional Server 1yr Subscription E-LTU

R1D32AAE

HPE Data Management Framework/Extents File System Additional Server 3yr Subscription E-LTU

R1D33AAE

HPE Data Management Framework/Extents File System Additional Server 5yr Subscription E-LTU

R1D34AAE

Ordering Guidelines:

1. Any one of the Base SKUs can be ordered standalone for new installations.

2. The Shared File System Base license covers 1 Metadata server and up to 511 SAN clients for the selected subscription term.

3. The Suite Base license covers HPE XFS, XVM, and R-pool in non-clustered mode running on a single server.

4. License terms (1, 3, 5 year) can NOT be stacked. For example, a 1-year license cannot be simultaneously stacked with a 3-year license to have 4 years of subscription.

5. Add one Increment SKU for each additional Metadata server configured.

6. Increment SKUs must have the same subscription term as the Base SKU.

Example: 1 Metadata server with any number of client nodes for a 1 year subscription

Qty

HPE SKU

Description

1

R1D38AAE

HPE Extents File System 1-server Base System 1yr Subscription E-LTU

Example: 1 Metadata server with any number of client nodes for a 5 year subscription

Qty

HPE SKU

Description

1

R1D40AAE

HPE Extents File System 1-server Base System 5yr Subscription E-LTU

For High Availability, configure at least 2 Metadata Servers.

Example: 2 Metadata servers with any number of client nodes for a 3 year subscription

Qty

HPE SKU

Description

1

R1D39AAE

HPE Extents File System 1-server Base System 3yr Subscription E-LTU

1

R1D33AAE

HPE Data Management Framework/Extents File System Additional Server 3yr Subscription E-LTU

Supported Operating Systems

Metadata Servers

• RHEL 7.9
• SLES 12 SP5 

NAS Edge Servers

• RHEL 7.9
• RHEL 8.4
• SLES 12 SP5 

Client Nodes

RHEL x86_64

• RHEL 7.1 to 7.9
• RHEL 8.0 to 8.4 

SLES x86_64

• SLES 12 SP2 to SP5
• SLES 15 GA to SP3

Non-clustered HPE XFS and XVM

RHEL x86_64

• RHEL 7.9
• RHEL 8.0 to 8.4 

SLES x86_64

• SLES 12 SP5
• SLES 15 GA to SP3

Maximum Cluster Size

512 total nodes including up to 3 Metadata Servers

Supported SAN Fabrics

• Fibre Channel
• InfiniBand (SRP and iSER)

HSM API Support

DMAPI Version 3

HPE Warranty

Worldwide, 24-hour support is available for the HPE Clustered Extents File System software solution through HPE. Customers are provided with complete installation services and given access to advanced services such as proactive health checks and migration assistance.

Service and Support

• 1 year 7x24 (Foundation Care)
• 3 year 7x24 (Foundation Care)
• 5 year 7x24 (Foundation Care)

Technology

Journaled, shared, 64-bit with file system consistency

Maximum file size

9 million terabytes

Maximum filesystem size

18 million terabytes

File system block size

• Selectable at file system creation time
• 512 bytes to 64 KB for normal data and up to 1 MB for real-time data
• File system extents (contiguous data) are configurable at file creation time using fcntl and are multiples of the file system block size; single extent can be up to 2GiB in size

Partitioning

Up to 64,000 volume elements, 64,000 wide stripes and dynamic configuration

Physical disk sector size supported

512 bytes and 4K Advanced Format

Backup/restore

• xfsdump and xfsrestore
• Linux Dump/Restore tools including bru, cpio, and tar
• Atempo Time Navigator, EMC NetWorker, and many popular commercial backup packages
• Dumps of active filesystems are supported

Support for data lifecycle management policy automation

The data management API (DMIG-DMAPI) allows implementation of data lifecycle management policy automation software without any kernel modifications. HPE Data Management Framework (DMF) is available to enable HSM Management of HPE XFS filesystems.

Date

Version History

Action

Description of Change

04-Oct-2021

Version 5

Changed

Updated for DMF Suite 4.5 release.

06-Jul-2020

Version 4

Changed

Updated for ISSP4.3 release.

Configuration Information and Additional Options sections were updated.

04-Nov-2019

Version 3

Changed

Updated for ISSP4.2 release and DMF Suite SKUs

Overview, Standard Features, Configuration Information, Additional Options and Technical Specifications sections were updated.

01-Oct-2018

Version 2

Changed

Updated for ISSP4 release

Overview, Standard Features, Configuration Information, Additional Options and Technical Specifications sections were updated.

07-Aug-2017

Version 1

New

New QuickSpecs

© Copyright 2021 Hewlett Packard Enterprise Development LP. The information contained herein is subject to change without notice. The only warranties for Hewlett Packard Enterprise products and services are set forth in the express warranty statements accompanying such products and services. Nothing herein should be construed as constituting an additional warranty. Hewlett Packard Enterprise shall not be liable for technical or editorial errors or omissions contained herein.  

Microsoft and Windows NT are US registered trademarks of Microsoft Corporation.  

Intel, the Intel logo, Xeon and Xeon Inside are trademarks of Intel Corporation in the U.S. and other countries. 

Linux is a registered trademark of Linus Torvalds.  

SUSE is a registered trademark of Suse. Ubuntu and Canonical are registered trademarks of Canonical Ltd.  

Red Hat is a trademark of Red Hat, Inc. in the U.S. and other countries.  

VMware is a registered trademark of VMware, Inc. in the United States and/or other jurisdictions. 

a00018442enw - 15994 - WorldWide  - V5 - 04-October-2021