1. IBM Spectrum Scale (formerly GPFS)
A cluster file system with high-performance, high availability and parallel file access
Elastic Storage V1 will be supported on Linux on System z. It will enable enterprise clients to use a high available cluster file system with Linux on System z in LPAR or Linux on z/VM.
IBM and ISV solutions (e.g. WebSphere MQ or Tivoli Storage Manager) will provide higher value for Linux on System z clients by exploiting Elastic Storage functionality as described below.
Client Value:
A highly available cluster architecture
Improved data availability through data access even when the cluster experiences storage or node malfunctions
Capabilities for high-performance parallel workloads
Concurrent high-speed, reliable file access from multiple nodes in the cluster environment
Smooth, non-disruptive capacity expansion and reduction
Services to effectively manage large and growing quantities of unstructured data
05/14/ Correct branding usage for Elastic Storage
* We have announced new code name: Elastic Storage software.
* We are not using 'IBM' in front of 'Elastic Storage' as this is not a formally named offering
* In general, only use "code name: Elastic Storage" in the first usage, and then simply 'Elastic Storage' in subsequent communication
* Never use "based on GPFS technology" in a slide title. It can be as a bullet on a page, or more ideally, in a footnote
* We should not be using "GPFS" as an offering name

2. Clustered and Distributed File Systems
Clustered file systems
File system shared by being simultaneously mounted on multiple servers accessing the same storage
Examples: IBM Spectrum Scale, Oracle Cluster File System (OCFS2), Global File System (GFS2), Lustre
Distributed file systems
File system is accessed through a network protocol and do not share block level access to the same storage
Examples: NFS, OpenAFS, CIFS
A clustered file system is a file system which is shared by being simultaneously mounted on multiple servers. There are several approaches to clustering, most of which do not employ a clustered file system (only direct attached storage for each node). Clustered file systems can provide features like location-independent addressing and redundancy which improve reliability or reduce the complexity of the other parts of the cluster. Parallel file systems are a type of clustered file system that spread data across multiple storage nodes, usually for redundancy or performance.
Key differences vs. distributed files systems:
Cluster file systems provide access from multiple nodes over a SAN
Direct access from the computer using a SAN provides performance advantages over standard network protocols.
Cluster file systems are tightly coupled and communicate at a more sophisticated level to enable an application, e.g. to have multiple nodes reading and writing to a single file.
OCFS2: Oracle Cluster File System 2 is a general-purpose shared-disk cluster file system from Oracle for Linux capable of providing both high performance and high availability. The file system is currently being used in virtualization (Oracle VM) in both the management domain, to host virtual machine images, and in the guest domain, to allow Linux guests to share a file system. It is also being used in database clusters (Oracle RAC), middleware clusters (Oracle E-Business Suite), etc.
GFS2: Global File System 2 is a shared disk file system for Linux computer clusters. GFS2 differs from distributed file systems (such as AFS) because GFS2 allows all nodes to have direct concurrent access to the same shared block storage.
NFS: Network file system is a protocol developed by Sun Microsystems which allows the access to files in a network. The data is not transferred, the user can access the data as if they were stored on the local hard disk.
OpenAFS: Open-Source-implementation of the Andrew File System offers a client-server architecture for federated file sharing and replicated read-only content distribution, providing location independence, scalability, security, and transparent migration capabilities. AFS is available for a broad range of heterogeneous systems including UNIX, Linux, and Windows.
CIFS: Common Internet File System is a network protocol for the Remote File-Access. It is the successor to the SMB protocol and is supported by most Windows servers and many other commercial servers and Network Attached Storage appliances as well as by the popular Open Source server Samba.
IBM Spectrum Scale - General overview

3. Virtualization Management
IBM Spectrum Scale
IBM’s shared disk, parallel cluster file system
Hardware resources
Virtualization Management
Linux
Cluster: Fast reliable communication, common admin domain
Shared disk: all data and metadata on storage devices accessible from any node through block I/O interface (“disk”: any kind of block storage device)
Parallel: data and metadata flow from all of the nodes to all of the disks in parallel.
Features: Native encryption, compression, native protocols, cloud tiering, disaster recovery, native raid
A cluster file system can be accessed from many computers at the same time over a network or a SAN.
The file system is built from a collection of arrays which contain the file system data and metadata. A file system can be built from a single disk or contain thousands of disks storing petabytes of data. Each file system is accessible from all nodes in the cluster.
Applications can access files through standard file system interfaces or through enhanced interfaces available for parallel programs. Applications can concurrently read or update a common file from multiple nodes in the cluster. GPFS maintains the coherency and consistency of the file system using sophisticated locking management and logging tools.
Elastic Storage provides unparalleled performance for large data objects and for large aggregates of smaller objects. It achieves high performance I/O by:
Striping data across multiple disks attached to multiple nodes.
High performance metadata (inode) scans.
Supporting a wide range of file system block sizes, configurable by the administrator, to match I/O requirements.
Utilizing advanced algorithms that improve read-ahead and write-behind IO operations. Using block level locking based on a very sophisticated scalable token management system to provide data consistency while allowing multiple application nodes concurrent access to the files.
All data stored in the file system is stripped across all of the disks within a storage pool. This wide data striping allows for optimal performance for the available storage.
GPFS provides unmatched performance and reliability with scalable access to critical file data. IT distinguishes itself from other cluster file systems by providing concurrent high-speed file access to applications executing on multiple nodes of a homogenous or heterogeneous cluster.

4. IBM Spectrum Scale - The Industry Standard for High Performance, Scalable Storage
Over 17 years in the marketplace
Broad adoption across many industries
Technical computing - Government, Educational
Enterprise computing - Business Analytics, Financial Services, Electronic Design Automation, Life Sciences, Oil & Gas Exploration, Media
Over 3,000 customers world-wide
Growing in double digits year to year
Industry proven, minimizing risk

5. Spectrum Scale Pioneered Big Data Management
File system
264 files per file system
Maximum file system size: 299 bytes
Maximum file size equals file system size
Customer with 25 PB file system
Nodes
1 to 16,384 Nodes
Extreme Scalability
Performance
High Performance Metadata
Striped Data
Parallel file access protocol
Integrated Tiered storage
Commodity hardware
Proven Reliability
Shipping since 1998
Built-in cluster availability automatic failover/failback
Data replication
Add/remove on the fly
Nodes
Storage/Disks
Rolling Upgrades
Administer from any node 5

6. Spectrum Scale Enables An Extremely Flexible Architecture
Tokyo NY
100 = TB
100 = TB
20 = PB
50 = PB
Case 1
2 servers with Spectrum Scale Software
Fronting 100 TB storage
Case 2: Add Performance
Add compute nodes
Faster Network
Case 3: Increase Capacity
Add any storage
Virtually unlimited scaling
Case 4: Global Share
Use Active File Management to expand your global namespace
Now, let’s play with the arch a bit. This will demonstrate the flexibility of a software defined storage architecture.
Adding new devices (disks or nodes) has no negative impact on data mapping, and new drives and nodes can be added without creating access bottlenecks or hot spots, making system management that much easier.
IBM and Business Partner Use Only

7. Spectrum Scale Features & Applications
Standard file system interface with POSIX semantics
Metadata on shared storage
Distributed locking for read/write semantics
Highly scalable
High capacity (up to 299 bytes file system size, up to 263 files per file system)
High throughput (TB/s)
Wide striping
Large block size (up to 16MB)
Multiple nodes write in parallel
Advanced data management
Snapshots, storage pools, ILM (filesets, policy)
Backup HSM (DMAPI)
Remote replication, WAN caching
High availability
Fault tolerance (node, disk failures)
On-line system management (add/remove nodes, disks, ...)
Hardware resources
Virtualization Management
Linux
Standard file system interface with POSIX semantics
Metadata on shared storage
Same on-disk structure as regular files, but...not directly visible to the user (don’t appear in any directory)
Distributed locking for read/write semantics
Distributed locking to synchronize updates to file system metadata on disk to prevent corruption, maintain cache consistency of data and metadata cached in memory on different nodes.
Highly scalable
High capacity (up to 299 bytes file system size, up to 263 files per file system)
High throughput (TB/s)
Wide striping
Large block size (up to 16MB)
Multiple nodes write in parallel
Advanced data management
Elastic Storage can help to achieve data lifecycle management efficiencies through policy-driven automation and tiered storage management. The use of storage pools, filesets and user-defined policies provide the ability to better match the cost of the storage to the value of the data. Storage pools allow to manage groups of disks within a file system. Using storage pools allows to create tiers of storage by grouping disks based on performance, locality or reliability characteristics.
Snapshots, storage pools, Information Life-cycle Management (file sets, policy):
Not all storage is the same: some is faster, cheaper, more reliable; Not all data are the same: some are more valuable, important, popular, …
Snapshot: Logical, read-only copy of the file system at a point in time. Typical uses: Backup: obtain consistent state of the file system, On-line access to previous file system state
Fileset: A partition of the file system name space (sub-directory tree); Allows administrative operations at finer granularity than entire file system, e.g., disk space limits, user/group quota, snapshots, caching, ...
Storage Pool: A named collection of disks with similar attributes intended to hold similar data
Policy: A set of user-specified rules that match data to the appropriate pool. E.g.: a migration policy can move data between pools, changes replication, delete data, or run arbitrary user commands. Automated, policy-driven storage management can reduce storage costs up to 90 percent.
Backup HSM (DMAPI) (HSM - Hierachical Storage Manager; DMAPI - Data Management API): External storage managed by Data Management Application (e.g., Tivoli HSM) that interacts with the file system through a standardized interface (DMAPI)
Move infrequently accessed data to external storage (e.g., tape)
Restore transparently on-demand when needed
Remote replication, WAN caching: Enable sharing file systems between nodes in different clusters, suitable for less reliable, higher-latency networks
High availability
Elastic Storage is fault tolerant and can be configured for continued access to data even if cluster nodes or storage systems fail. This is accomplished though robust clustering features and support for data replication. Elastic Storage includes the complete set of node availability components to handle data consistency and availability. In an Elastic Storage cluster all nodes see all data and all cluster operations can be done by any node in the cluster with a server license. This means that there are no special nodes in an Elastic Storage cluster. All nodes are capable of performing all tasks.
The file system can be configured that it remains available automatically if a disk or server fails. Elastic Storage can be configured to automatically recover from node, storage and other infrastructure failures.
Fault tolerance (node, disk failures)
On-line system management (add/remove nodes, disks, ...)

8. Flexible Topologies for GPFS Cluster Configuration
Network Shared Disk (NSD) Server Model
Shared Nothing Cluster Model
Storage Area Network (SAN) Model
Storage
TCP/IP or Infinband RDMA Network
Storage Network
TCP/IP Network
TCP/IP or Infinband Network
I/O Servers
Application Nodes
GPFS supports 3 topologies – SAN, NSD, and SNC
In all three configurations, all the features of the global file system are maintained
– global name space and sharing
Software enables this flexibility !!!! 8

9. What Spectrum Scale is NOT
Not a client-server file system like NFS, CIFS or AFS
TCP/IP Network
File Server
Client Nodes
Storage
No single-server performance and bottleneck scaling limits
data
data
metadata
Network
Metadata
Server
No single-server performance and bottleneck scaling limits
When large numbers of clients want to access the data, or if the data set grows too large, a NFS server quickly becomes the bottleneck and significantly impacts system performance because the NFS server sits in the data path between the client computer and the physical storage devices.
An Elastic Storage file system can be built from a single disk or contain thousands of disks storing petabytes of data. Each file system is accessible from all nodes in the cluster.
No centralized metadata server
Elastic Storage provides scalable metadata management by allowing all nodes of the cluster accessing the file system to perform file metadata operations. This key and unique feature distinguishes Elastic Storage from other cluster file systems which typically have a centralized metadata server handling fixed regions of the file namespace. A centralized metadata server can often become a performance bottleneck for metadata intensive operations and can represent a single point of failure. Elastic Storage solves this problem by managing metadata at the node which is using the file or in the case of concurrent access to the file, at a dynamically selected node which is using the file.
Metadata
No centralized metadata server
Data
Data

10. The Logical File System View
IBM Spectrum Scale
Simple. Powerful. Economical.
Maximum file system size of one million yottabytes
. . .
One big file system or you can divide into as many as 256 smaller file systems
Each file system can be further divided into fileset containers (tree branches)
A rule can apply to any file being created or only to files being created within a specific fileset or group of filesets.
Define soft or hard quota by user, group or fileset
FS1
FS2
FS256
Spectrum Scale policy based data migration
File1
File1
File1
File2
File2
File2
File3
File3
File3
File3
File4
IBM and Business Partner Use Only

11. Enhanced protocol support
The challenge: How can I share my storage infrastructure across all of my legacy and new generation applications?
The Solution
The new IBM Spectrum Scale Protocol Node allows access to data stored in a Spectrum Scale filesystem, using additional access methods and protocols.
The Protocol Node functions are clustered and can support transparent failover for NFS and SWIFT protocols as well as SMB protocols.
Multiprotocol data access from other systems using the following protocols
NFS v3 and v4
SMB 2 and SMB 3.0 mandatory features / CIFS for Windows support.
OpenStack Swift and S3 API support for object storage.
Our next theme for this launch is enhanced protocol support.
The challenge: How can I share my storage infra across both legacy and new gen apps.
The solution: Our new Spectrum Scale protocol node now allows access to shared storage from Linux and AIX systems (NFS v3 and v4), Windows (SMB 2 and SMB 3) and object access via OpenStack Swift.

12. The Solution: IBM Spectrum Scale™ brings it all together
IBM Spectrum Scale™ replaces HDFS and NAS file storage
Full Hadoop interfaces for Map/Reduce analytics processing
No transfer or ingest required as the data is already there
Fully protected with Backup Software
File-level access support for NFS, CIFS, FTP, SCP and HTTPS
Supports Enterprise File Sync-and-Share via OwnCloud or Funambol
IBM Spectrum Scale™ replaces SAN-based file systems
Replaces NTFS, EXT4, JFS2 and other POSIX file systems
Used by over 200 of the top 500 supercomputers
No file transfers required between different OS
Can be used with everything from databases to video streaming
For x86, POWER and z System servers
Secure with Data-at-rest encryption
IBM Spectrum Scale™ offers Object access
Object-level access based on OpenStack Swift driver and Amazon S3 APIs
Global Name Space
IBM Spectrum Scale™ supports all media
Spans flash, disk and tape media

13. Unleash New Storage Economics on a Global Scale
Client workstations
Users and applications
Compute Farm
Single name space
SMB/CIFS
OpenStack
POSIX
Map Reduce Connector
Site A
Site B
Site C
Cinder
Swift
NFS
Manila
Glance
IBM Spectrum Scale Automated data placement and data migration
Off Premise
Tape
Flash
Storage Rich Servers
Multi-cloud Storage Toolkit
Disk

14. Backup

15. Why is Spectrum Scale of interest to customers
Why is Spectrum Scale of interest to customers? What problems does Spectrum Scale solve?
Insufficient Capacity and Performance
Spectrum Scale-based file servers can scale to enormous performance and capacity, avoiding storage islands and staying easy to manage.
Unreliable Storage
Spectrum Scale based file systems can survive failures of many components without incurring data loss and while remaining available. It has techniques for monitoring components and recovering from failures extremely quickly.
Cost Escalation
By avoiding storage islands much cost is saved. Management cost, cost of application downtime, cost of over/under provisioning storage, all can be reduced or eliminated. Additional cost can be saved by having Spectrum Scale automatically move files to cheaper disks or even to tape.
Geographically distributed data sharing
Spectrum Scale lets organizations securely share data across different branches and locations, providing remote access and automated distribution getting maximum value from data organization-wide.
IBM Spectrum Scale - General overview

19. Supports a Wide Range of Hardware and Software
Operating systems
Hardware platforms
Linux®
Red Hat Enterprise Linux v6 / 7
SUSE Linux Enterprise Server v11/12
Debian v6 / 7
Ubuntu 14.04/16.04
IBM AIX® v6.1 / 7.1
Windows® Server 2008 x64 (SP2); Windows® Server 2008 R2; Windows 7 x64 SP1; Windows Server 2012 R2; Windows 8.1
IBM Power Big Endian
IBM Power Little Endian
x86_64
z Systems
Including software and 3rd party application software support. IBM takes a comprehensive view of IT operations. One of the primary advantages of working with IBM is an integrated approach to your Elastic Storage solution.
IBM Scale Out Network Attached Storage
IBM Smart Business Storage Cloud
IBM Information Archive
IBM Storwize V7000 Unified Storage
IBM Smart Analytics System
IBM DB2 pureScale
IBM Systems for SAP HANA
IBM SmartCloud Enterprise
IBM SmartCloud Archive
IBM Digital Media Center
Information Life cycle management
SAP BI Accelerator
SAP/Business Objects
Oracle
SAS
Ab Initio
Informatica
SAF
...
Storage
IBM Storage and storage hardware from all vendors such as: EMC, Hitachi, Hewlett Packard, DDN
IBM Spectrum Scale - General overview 19

20. IBM Spectrum Scale benefits over NAS
Better performance
Eliminate hotspots with massively parallel access to files 
Sequential I/O with ES greater than 400 GB/s
Throughput advantage for parallel streaming workloads, e.g. Tech Computing and Analytics
More Storage. More Files. Hyper Scale.
Simplified Management
Easier management with one global namespace instead of managing islands of NAS arrays, e.g. no need to copy data between compute clusters
Integrated policy driven automation
Fewer storage administrators required
Lower Cost
Optimizes storage tiers including flash, disk and tape
Increased efficiency and more efficient provisioning due to parallelization and striping technology
Remove duplicate copies of data, e.g. run analytics on one copy of data without having to set up a separate silo

21. Object Store vs. File System
Interface
Web-based: GET/PUT/DELETE
RESTful: Stateless
Metadata
Synchronization
Eventual Consistency
No distributed locking
Software Defined Storage
Commodity hardware
Designed to fault but never fail
Built to auto-recover by design
Features
Basic services that scale (KISS)
SW extendible with web interfaces
Interface
POSIX: Open/Seek/Read/Write/Close
Stateful
Synchronization
Strict Consistency
Uses distributed locking
Hardware and Software
Best of breed hardware
Designed not to fault
Admin controlled recovery
Features
Abundant enterprise features built into the products
REST: Representational State Transfer, a Web Service API

22. The History of Spectrum Scale
This infographic is the genealogy of IBM Spectrum Scale, from it’s birth as a digital media server and HPC research project to it’s place as a foundational element in the IBM Spectrum Storage family. It highlights key milestones in the product history, usage, and industry to convey that Spectrum Scale may have started as GPFS, but it is so much more now. IBM has invested in the enterprise features that make it easy to use, reliable and suitable for mission critical storage of all types.
This marked a major milestone in a long history of development.
Those core features mentioned in the Press Release demonstrate the commitment IBM has to moving software-defined storage into the center of the enterprise. And, the pace of change is accelerating.
Footnote goes here

23. Resources ibm.com: Public Wiki: IBM Knowledge Center:
ibm.com/systems/platformcomputing/products/gpfs/
Public Wiki:
ibm.com/developerworks/community/wikis/home?lang=en#!/wiki/General Parallel File System (GPFS)
IBM Knowledge Center:
ibm.com/support/knowledgecenter/SSFKCN/gpfs_welcome.html?lang=en
Data sheet: IBM General Parallel File System (GPFS) Version 4.1
ibm.com/common/ssi/cgi-bin/ssialias?subtype=SP&infotype=PM&appname=STGE_DC_ZQ_USEN&htmlfid=DCD12374USEN&attachment=DCD12374USEN.PDF
Presentation: will be published soon
Spectrum Scale Quick Install for Linux on IBM System z
IBM Internal:
Spectrum Scale Sales Wiki: Sales Wiki/page/Elastic Storage Sales Wiki
GPFS Quick Reference Guide: bin/ssialias?subtype=ST&infotype=SA&appname=STGI_DC_ZQ_USEN&htmlfid=DCY12364USEN&attachment=DCY12364USEN.PDF
GPFS Conversation Starter: bin/ssialias?subtype=RG&infotype=PM&appname=STGI_DC_ZQ_USEN&htmlfid=DCO01637USEN&attachment=DCO01637USEN.PPT
Presentation: Selling General Parallel File System (GPFS™): bin/ssialias?subtype=PS&infotype=SA&appname=STGI_DC_ZQ_USEN&htmlfid=DCP03210USEN&attachment=DCP03210USEN.PPT
01.sso.ibm.com/learning/registry/assets/LearningTechnologies/ltu43563?sourceUrl=http%3A%2F%2Flt.be.ibm.com%2Fstg%2Fltu43563
Technical overview of GPFS - GPFS Tutorial 2013 (presentations from Frank Schmuck):
IBM Internal Only

24. Easier to Try Spectrum Scale virtual machine
Turn-key Spectrum Scale VM available for download
Try the latest Spectrum Scale enhancements
Full functionality on laptop, desktop or server
Incorporate external storage
Use for live demonstrations, proof of concepts, education, validate application interoperability
Scripted demonstrations
External download link
Download VM image, quick start guide, explorer guide, advanced user guide
Limitations
VirtualBox hypervisor only
Type-2 Hypervisor limits performance
Not supported for production workloads
Can not be migrated to bare metal
Easier to try. This release will also feature a completely functional virtual machine as a complementary download. Designed specifically to install on a laptop, desktop or server, you can see the latest UI, try out the unified file & object storage, test HDFS transparency or just run the smallest little version
Limited Beta today:
Internal Link:
GA 11/20: