In this slidecast, Jim Gutowski and Scott Fadden from IBM describe the advantages IBM Spectrum Scale Storage brings to the enterprise.
IBM Spectrum Scale is a proven, scalable, high-performance data and file management solution (based upon IBM General Parallel File System or GPFS, also formerly known as code name Elastic Storage). IBM Spectrum Scale provides world-class storage management with extreme scalability, flash accelerated performance, and automatic policy-based storage tiering from flash through disk to tape. IBM Spectrum Scale reduces storage costs up to 90% while improving security and management efficiency in cloud, big data & analytics environments."
Watch the video presentation starting 4/8/2015: http://wp.me/p3RLHQ-e0a
Learn more: http://ibm.co/1y1QIrj
RICH: Hi and welcome! This is the “Rich Report” from InsideBigData – and with me today is Jim Gutowski and Scott Fadden from IBM. Welcome guys!
JIM: Hi Rich – thanks for having us on the Rich Report!
RICH: So, audience – you may have heard about IBM’s recent introduction of the IBM Spectrum Storage family (solution?) – and if you listened to my Rich Report last week, you heard Jay Muelhoefer talk about Software Defined Infrastructure. Well, today, Jim & Scott are going to help us drill down into more detail on IBM Spectrum Scale – which was formerly IBM GPFS (also formerly referred to as Elastic Storage).
Rich Q: So Jim – I think we’re all aware of GPFS, but can you explain in a sentence how IBM Spectrum Scale is taking storage to the next level?
JIM: Hi Rich - Sure…IBM Spectrum Scale is the latest revision of GPFS – and it’s been more than an HPC solution for a long time! But, most of our customers who speak on our behalf are HPC customers – but what many may not know is that we have many clients and users in the Big Data space – such as Nuance Communications, CitiBank, and Cypress Semiconductor – who have been using our solution for years!
Jim talks: set the stage for what is happening in the industry today – challenges customers face around Data Explosion, Economics, and Innovation
“And now, I’ll turn it over to Scott to explain the Spectrum Scale architecture”….
SCOTT: Thanks Jim! [Scott give elevator pitch on Spectrum Scale]
RICH Q: For large scale HPC, the choice of parallel filesystems seems to always come down to GPFS and Lustre. What are the advantages of GPFS/Spectrum Scale?
IBM A: There are other customers who need ILM, integration with tape, tying {....} together - Lustre cannot do that
RICH Q: Can Spectrum Scale software be used with existing storage infrastructure from other vendors?
IBM A:
RICH Q: Does your business model lend itself towards other vendors selling Spectrum Scale on their storage hardware as part of a turnkey solution?
IBM A:
RICH Q: What are the forces driving technical computing towards object storage? (GUYS – you decide if / how you want to answer)
Background notes for this slide:
Spectrum Scale delivers High Bandwidth for Large Files
Parallel access from all clients to all storage: end to end
Parallel NSD protocol enabling Application node(s) to simultaneously access all the NSDs
High Bandwidth access to file(s)
Wide striping
Large block size
Parallel access to files from multiple nodes
Efficient deep pre-fetching: read ahead, write behind
Highly multi-threaded daemon
400GB/s to single file-system
Spectrum Scale delivers High file-system IOPS for Small Files and Metadata
High file-system IOPS
Distributed Metadata Architecture
Option to configure dedicated fast disks to store file-system metadata
Data for small files stored within the inode thereby reducing disk I/O for small file workloads.
Integrated Tiered storage
Traditional Storage: Small files require two I/O operations
metadata(inode)
file-data
Spectrum Scale Optimization – Access small files with ONE I/O operation
Allow data for small files and directories to be stored in the inode. This has the potential advantages of speeding up directory scans and reducing disk I/O for small file workloads. Supported Inode-Size: 512B, 1KiB, 4KiB
SCOTT OR JIM: Talk to Support, Services (instant), and proven track record from IBM.
Lustre has been un-supported and/or semi-supported by multiple vendors over the years - and what happens when you have a problem with that feature?
RICH: Provide positive observation comment before closing.