Expanding Adoption and Access to Next Generation of HPC Storage Utilizing Intel® Xeon® E5 Architecture
1 like494 views
The document discusses the adoption of next-generation HPC storage solutions powered by Intel E5 architecture, emphasizing enhanced performance and scalability for high-resolution simulations and industrial modeling. It highlights the benefits of Xyratex's ClusterStor storage systems, which offer superior reliability, availability, and serviceability for massive data workloads. Key features include high throughput, dynamic scaling, and integrated management, aiming to overcome legacy storage bottlenecks in HPC environments.
Expanding Adoption and Access to Next Generation of HPC Storage Utilizing Intel® Xeon® E5 Architecture
- 1. Expanding Adoption and Access to Next Generation HPC Storage Utilizing Intel E5 Architecture Michael K. Connolly Partner and Alliances DevelopmentPartner and Alliances Development
- 2. Commercial Scientific and Engineering Computing High resolution simulation and modeling of Everything •Exploding data capture, analysis and application performance scale •Time critical commercial simulation and industrial modeling requires • Faster job completion to attain significantly higher iteration rates • More precise application insights with higher resolution accuracy • Enhanced data throughput efficiency to overcome legacy barriers • Scale beyond NFS (not suited to intensive simulation needs) 2
- 3. •Massively parallel globally concurrent access • Thousands application clients - billions of files •Highest performance scalability, reliability, availability and serviceability • Single file system namespace in 10s High Performance Applications • Single file system namespace in 10s petabytes usable capacity • Linear file system scale-out performance in 100s of GB/s •Integrated solution management • End-to-end system visibility, provisioning, manageability and diagnostics
- 4. OEM Storage Solutions HDD Capital Equipment Solutions Ultra-dense, high performance and high- availability data storage enclosure & integrated Fully automated disk drive test solutions, servo-track writing technology and disk processing solutions for high-volume manufacturing environments. ClusterStor™ Storage Solutions Fully integrated scale-out file system solutions engineered to meet HPC requirements in research, government & commercial applications. enclosure & integrated application platforms.
- 5. Leading Provider of Data Storage Solutions HDD Capital Equipment Solutions ~ 50% of w/w disk drives are produced utilizing Xyratex Technology* Largest independent supplier of Disk Drive Capital Equipment OEM Storage Solutions 15%of w/w external storage capacity shipped in 2012* > 3,000 Petabytes of storage shipped in 2012 Largest OEM Disk Storage System provider Capital EquipmentStorage System provider ClusterStor™ HPC Storage Solutions Fastest Data Storage System in the World ~ 90PB ClusterStor capacity shipped in 2012 Powering research, energy, defense and life science applications worldwide *Source: Company estimates
- 6. Integrated File System and Appliance Solutions Embedded Storage Platforms Storage Application Platform Modular Enclosures Storage Enclosures Enterprise Storage Solutions: OneStor™ and ClusterStor™ Integrated Application Platforms Integrated System Solutions Storage Application Platform Xyratex Storage Innovations OneStor™ Software (USM) Common Mgmt SW & Deep System Diagnostics Integrated System Testing Proprietary Pre-Ship Testing Methods & Automation Metis™ Power Protection UPS Alternative for Data Protection Reliability, Availability & Serviceability Enterprise Redundancy & Resiliency
- 7. Delivering Ultimate Scale-Out Storage for HPC and Big Data Optimized time to productivity Efficiency, application availability, results ClusterStor™ 6000 Industry Leading Performance at Massive Scale Unmatched file system performance – Delivered! Industry’s fastest just got two times faster Highest reliability, availability and serviceability Enterprise level resiliency
- 8. ClusterStor – An Integrated Lustre® Storage Solution Engineered for High Performance Computing ClusterStor™ consolidates: controllers, storage, operating system, and the Lustre file system creating a new optimal scale-out storage Fully realized solution provides the industry’s best reliability, scalability, superior performance and ease of management. • Performance • Scale-Out Capability • Capacity and Density Scaling • Ease of Management • Ease of Deployment
- 9. ClusterStor - Dynamic Scaling Traditional Scaling Approach Network I/O Ports: Dynamic Scaling Approach (SSU) Overall Rack Amounts Compute & RAM: Balanced Performance & ScalabilityUnbalanced Performance Total HDDs: Compute & RAM: <25GB/sec File System Throughput = CS-6000 >42GB/sec File System Throughput =
- 10. ClusterStorTM Scalable Storage Unit (SSU) State of the Art Technology Highlights Xyratex uses Intel Xeon® E5 processor family to power ClusterStorTM 6000 • High-performance per watt processing power • Advanced Storage Features • Data Protection and Security • Intel Integrated I/O • Scalability• Scalability ClusterStorTM 6000 Scalable Storage Unit provides 6GB/s file system throughput in 5U high availability scale-out building block • Expanded memory capacity, wider and faster memory bandwidth • PCIe Gen3 faster and wider with more lanes, higher IO throughput • FDR InfiniBand, lower latency, higher IO throughput • No compromise high quality, high integrity Near Line (NL) SAS HDDs
- 11. ClusterStor™ Delivering the Ultimate in HPC Data Storage Faster Data Throughput & Capacity Increased Compute Capacity Enablement Optimized time 11 Higher Levels of System Availability Efficient Systems Deployment & Management Optimized time to productivity
- 12. Integrated System Solution Management Key Features • Remediation Center Intelligence-guided remediation steps • Expert system Built into service oriented architecture • Health monitoring telemetry Benefits • Reduced service cost to manage large scale HPC data storage • Modular end-to-end cluster-centric management solution • Extensible framework to integrate with customer environments • Reduced time to issue resolution, with automated real time cluster metric and inventory analysis assisted by expert system • Health monitoring telemetry database API Enables QoS analytics export and repurposing
- 13. •Massively parallel globally concurrent access • Thousands application clients - billions of files •Highest performance scalability, reliability, availability and serviceability • Single file system namespace in 10s High Performance Applications • Single file system namespace in 10s petabytes usable capacity • Linear file system scale-out performance in 100s of GB/s •Integrated solution management • End-to-end system visibility, provisioning, manageability and diagnostics
- 14. • Fundamentally overcomes the well known “Storage I/O Bottleneck” • Satisfies HPC and Big Data accelerated pace for more compute and capacity • Resolves legacy “cobbled together ClusterStor Benefits Increased Compute Capacity Faster Data Throughput & Capacity • Resolves legacy “cobbled together systems” reliability and availability issues • Delivers fastest time to productivity Higher Levels of System Availability Efficient Deployment & Management
- 15. ClusterStor Benefits Shrinking Budgets Limited Power Resources • Collapses and consolidates solution into modular, easy to use building blocks • Savings in storage hardware resulting in lower latencies, higher reliability and reduced power and cooling costs Datacenter Infrastructures reduced power and cooling costs • Reclaims resources for more compute, performance and storage capacity • Eliminates waste
- 16. Global Customer Relationships Strategic relationships established with both market & technology leaders
- 17. Strategic Alliances and Technology Partners Strategic Association Membership
- 18. Some ClusterStor Solution Success Ken Atkinson High Performance Computing, AWE “All of the performance testing met or exceeded expectations.” Jacques-Charles Lafoucriere Department Leader, CEA “The software stack provided with the system enabled us to install a working Lustre environment within hours.” Paul Calleja Director of High Performance Computing, University of Cambridge “With Lustre and the comprehensive storage management framework we were able to easily configure and manage our cluster”
- 19. Come see more of us at Booth #364 THANK YOU! Questions? Come see more of us at Booth #364 michael_connolly@xyratex.com www.xyratex.com Contact Us