This document provides information about SGI's HPC customers and systems. It summarizes installations including a 1.009 PFlop system at NASA, a 339 TFlop system in Germany, and a 155.7 TFlop system in Norway. It also outlines SGI's key markets like HPC, cloud storage, big data, and commercial scientific applications. In summary, SGI is a leader in technical computing supporting mission critical applications through large-scale systems for research, industry, and government.
In this slidecast, Bill Mannel from SGI presents an update on the company's innovative HPC solutions.
Learn more at: http://sgi.com
Watch the presentation video: http://insidehpc.com/2013/07/01/slidecast-sgi-product-update-for-june-2013/
Designed without compromise for unlimited innovation, Bull's HPC clusters of bullx processors are deployed on several continents with petascale computing power, for applications from sports car design to simulating the whole full observable universe.
In this deck from Switzerland HPC Conference, Gunter Roeth from NVIDIA presents: Deep Learning on the SaturnV Cluster.
"Machine Learning is among the most important developments in the history of computing. Deep learning is one of the fastest growing areas of machine learning and a hot topic in both academia and industry. It has dramatically improved the state-of-the-art in areas such as speech recognition, computer vision, predicting the activity of drug molecules, and many other machine learning tasks. The basic idea of deep learning is to automatically learn to represent data in multiple layers of increasing abstraction, thus helping to discover intricate structure in large datasets. NVIDIA has invested in SaturnV, a large GPU-accelerated cluster, (#28 on the November 2016 Top500 list) to support internal machine learning projects. After an introduction to deep learning on GPUs, we will address a selection of open questions programmers and users may face when using deep learning for their work on these clusters."
Watch the video: http://wp.me/p3RLHQ-gDv
Learn more: http://www.nvidia.com/object/dgx-saturnv.html
and
http://hpcadvisorycouncil.com/events/2017/swiss-workshop/agenda.php
Sign up for our insideHPC Newsletter: http://insidehpc.com/newsletter
The IBM POWER10 processor represents the 10th generation of the POWER family of enterprise computing engines. Its performance is a result of both powerful processing cores and high-bandwidth intra- and inter-chip interconnect. POWER10 systems can be configured with up to 16 processor chips and 1920 simultaneous threads of execution. Cross-system memory sharing, through the new Memory Inception technology, and 2 Petabytes of addressing space support an expansive memory system. The POWER10 processing core has been significantly enhanced over its POWER9 predecessor, including a doubling of vector units and the addition of an all-new matrix math engine. Throughput gains from POWER9 to POWER10 average 30% at the core level and three-fold at the socket level. Those gains can reach ten- or twenty-fold at the socket level for matrix-intensive computations.
From Rack scale computers to Warehouse scale computersRyousei Takano
This document discusses the transition from rack-scale computers to warehouse-scale computers through the disaggregation of technologies. It provides examples of rack-scale architectures like Open Compute Project and Intel Rack Scale Architecture. For warehouse-scale computers, it examines HP's The Machine project using application-specific cores, universal memory, and photonics fabric. It also outlines UC Berkeley's FireBox project utilizing 1 terabit/sec optical fibers, many-core systems-on-chip, and non-volatile memory modules connected via high-radix photonic switches.
This presentation covers a talk on the topic of "AI on the edge". The talk was delivered in the Conference on Artificial Intelligence and Robotics Technology held on Jan 28, 2021 by National Center of Artificial Intelligence Pakistan & working group by Ministry of Science and Technology on AI & Robotics.
In this deck, Gilad Shainer from Mellanox announces the world’s first HDR 200Gb/s data center interconnect solutions. "These 200Gb/s HDR InfiniBand solutions maintain Mellanox’s generation-ahead leadership while enabling customers and users to leverage an open, standards-based technology that maximizes application performance and scalability while minimizing overall data center total cost of ownership. Mellanox 200Gb/s HDR solutions will become generally available in 2017.
Watch the video presentation: http://insidehpc.com/2016/11/hdr-infiniband/
In this slidecast, Bill Mannel from SGI presents an update on the company's innovative HPC solutions.
Learn more at: http://sgi.com
Watch the presentation video: http://insidehpc.com/2013/07/01/slidecast-sgi-product-update-for-june-2013/
Designed without compromise for unlimited innovation, Bull's HPC clusters of bullx processors are deployed on several continents with petascale computing power, for applications from sports car design to simulating the whole full observable universe.
In this deck from Switzerland HPC Conference, Gunter Roeth from NVIDIA presents: Deep Learning on the SaturnV Cluster.
"Machine Learning is among the most important developments in the history of computing. Deep learning is one of the fastest growing areas of machine learning and a hot topic in both academia and industry. It has dramatically improved the state-of-the-art in areas such as speech recognition, computer vision, predicting the activity of drug molecules, and many other machine learning tasks. The basic idea of deep learning is to automatically learn to represent data in multiple layers of increasing abstraction, thus helping to discover intricate structure in large datasets. NVIDIA has invested in SaturnV, a large GPU-accelerated cluster, (#28 on the November 2016 Top500 list) to support internal machine learning projects. After an introduction to deep learning on GPUs, we will address a selection of open questions programmers and users may face when using deep learning for their work on these clusters."
Watch the video: http://wp.me/p3RLHQ-gDv
Learn more: http://www.nvidia.com/object/dgx-saturnv.html
and
http://hpcadvisorycouncil.com/events/2017/swiss-workshop/agenda.php
Sign up for our insideHPC Newsletter: http://insidehpc.com/newsletter
The IBM POWER10 processor represents the 10th generation of the POWER family of enterprise computing engines. Its performance is a result of both powerful processing cores and high-bandwidth intra- and inter-chip interconnect. POWER10 systems can be configured with up to 16 processor chips and 1920 simultaneous threads of execution. Cross-system memory sharing, through the new Memory Inception technology, and 2 Petabytes of addressing space support an expansive memory system. The POWER10 processing core has been significantly enhanced over its POWER9 predecessor, including a doubling of vector units and the addition of an all-new matrix math engine. Throughput gains from POWER9 to POWER10 average 30% at the core level and three-fold at the socket level. Those gains can reach ten- or twenty-fold at the socket level for matrix-intensive computations.
From Rack scale computers to Warehouse scale computersRyousei Takano
This document discusses the transition from rack-scale computers to warehouse-scale computers through the disaggregation of technologies. It provides examples of rack-scale architectures like Open Compute Project and Intel Rack Scale Architecture. For warehouse-scale computers, it examines HP's The Machine project using application-specific cores, universal memory, and photonics fabric. It also outlines UC Berkeley's FireBox project utilizing 1 terabit/sec optical fibers, many-core systems-on-chip, and non-volatile memory modules connected via high-radix photonic switches.
This presentation covers a talk on the topic of "AI on the edge". The talk was delivered in the Conference on Artificial Intelligence and Robotics Technology held on Jan 28, 2021 by National Center of Artificial Intelligence Pakistan & working group by Ministry of Science and Technology on AI & Robotics.
In this deck, Gilad Shainer from Mellanox announces the world’s first HDR 200Gb/s data center interconnect solutions. "These 200Gb/s HDR InfiniBand solutions maintain Mellanox’s generation-ahead leadership while enabling customers and users to leverage an open, standards-based technology that maximizes application performance and scalability while minimizing overall data center total cost of ownership. Mellanox 200Gb/s HDR solutions will become generally available in 2017.
Watch the video presentation: http://insidehpc.com/2016/11/hdr-infiniband/
The document discusses the path to exascale computing and the challenges involved. It outlines existing trends like the end of Moore's law and major technology challenges. Technologies being developed to overcome issues include more efficient interconnects, memory, storage, and processors. The development of exascale systems will also require rethinking architectures and a paradigm shift in priorities like power efficiency. Significant code modernization efforts will be needed to effectively utilize exascale systems and harness massively parallel computing.
Jax 2013 - Big Data and Personalised MedicineGaurav Kaul
Global healthcare trends are driving an increase in data and a need for personalized medicine approaches. These include rising healthcare costs as populations age, with the average global age expected to rise from 10% to 21% over 60 by 2050. Next generation sequencing is driving down costs, enabling large amounts of genomic and other health data to be collected. This presents big data challenges to manage and analyze the data to enable personalized medicine approaches by 2020. Intel is working on solutions across the hardware and software stack to help with big healthcare data challenges including high performance computing, optimized software frameworks, data analytics methods, and use case examples like sequencing appliances.
Supermicro designed and implemented a rack-level cluster solution for San Diego Supercomputing Center (SDSC) optimized for their custom and experimental AI training and inferencing workloads, and meeting their environmental and TCO requirements. The project team will discuss the journey of designing and deploying our Rack Plug and Play cluster, and Shawn Strande, Dupty Director, SDSC, will be sharing his experience of partnering with the Supermicro team to solve his challgenges in HPC and AI.
The team will also share the technology that powers the SDSC Voyager Supercomputer, the Habana Gaudi AI system with 3rd Gen Intel® Xeon® Scalable processors for Deep Learning Training, and Habana Goya for Inferencing.
Watch the webinar: https://www.brighttalk.com/webcast/17278/517013
This document provides an overview of a new CPU capability called Intel® Speed Select
Technology – Base Frequency (Intel® SST-BF), which is available on select SKUs of 2nd
generation Intel® Xeon® Scalable processor (formerly codenamed Cascade Lake). The
document also includes benchmarking data and instructions on how to enable the
capability.
Value propositions of this capability include:
• Select SKUs of 2nd generation Intel® Xeon® Scalable processor (5218N, 6230N, and
6252N) offer a new capability called Intel® SST-BF.
• Intel® SST-BF allows the CPU to be deployed with an asymmetric core frequency
configuration.
• The placement of key workloads on higher frequency Intel® SST-BF enabled cores
can result in an overall system workload increase and potential overall energy
savings when compared to deploying the CPU with symmetric core frequencies
Fujitsu World Tour 2017 - Compute Platform For The Digital WorldFujitsu India
Fujitsu has decades of experience designing and manufacturing servers. Their PRIMERGY servers are known for best-in-class quality that ensures continuous operation with almost no unplanned downtimes. This is achieved through rigorous testing and manufacturing processes in their state-of-the-art factories in Germany. Fujitsu's demand-driven manufacturing approach allows them to produce servers flexibly based on current orders, enabling fast response times and fulfilling individual customer requests.
AMD and the new “Zen” High Performance x86 Core at Hot Chips 28AMD
The document summarizes a presentation about AMD's new "Zen" x86 CPU core architecture. The Zen architecture provides a 40% increase in instructions per clock compared to previous cores through improvements in the core engine, caches, floating point capabilities, and the addition of simultaneous multithreading. The Zen core was designed from the ground up to optimize performance and power efficiency across applications from notebooks to supercomputers.
dCUDA: Distributed GPU Computing with Hardware Overlapinside-BigData.com
Torsten Hoefler from ETH Zurich presented this deck at the Switzerland HPC Conference.
"Over the last decade, CUDA and the underlying GPU hardware architecture have continuously gained popularity in various high-performance computing application domains such as climate modeling, computational chemistry, or machine learning. Despite this popularity, we lack a single coherent programming model for GPU clusters. We therefore introduce the dCUDA programming model, which implements device-side remote memory access with target notification. To hide instruction pipeline latencies, CUDA programs over-decompose the problem and over-subscribe the device by running many more threads than there are hardware execution units. Whenever a thread stalls, the hardware scheduler immediately proceeds with the execution of another thread ready for execution. This latency-hiding technique is key to make best use of the available hardware resources. With dCUDA, we apply latency hiding at cluster scale to automatically overlap computation and communication. Our benchmarks demonstrate perfect overlap for memory bandwidth-bound tasks and good overlap for compute-bound tasks."
Watch the video presentation: http://wp.me/p3RLHQ-gCB
1. The document discusses Microsoft's data center strategy and optimization efforts to improve performance per watt, dollar, and reduce total cost of ownership.
2. Key aspects of Microsoft's strategy include modular pre-fabricated data center components, increased server and data center efficiency through workload analysis and low power processors, and reducing infrastructure costs through techniques like high-temperature cooling.
3. Microsoft has invested over $500 million to build large, efficient data centers like its Chicago facility, and is working towards new designs with no mechanical cooling and ultra-modularity to further reduce costs and increase scalability.
Microsoft Project Olympus AI Accelerator Chassis (HGX-1)inside-BigData.com
In this video from the Open Compute Summit, Siamak Tavallaei from Microsoft presents an overview of the Microsoft Project Olympus AI Accelerator Chassis, also known as the HGX-1.
Watch the presentation video: http://wp.me/p3RLHQ-guX
Sign up for our insideHPC Newsletter: http://insidehpc.com/newsletter
The GIST AI-X Computing Cluster provides powerful accelerated computation resources for machine learning using GPUs and other hardware. It includes DGX A100 and DGX-1V nodes with 8 NVIDIA A100 or V100 GPUs each, connected by high-speed networking. The cluster uses Singularity containers, Slurm scheduling, and Ceph storage. It allows researchers to request resources, build container images, and run distributed deep learning jobs across multiple GPUs.
Evolution of Supermicro GPU Server SolutionNVIDIA Taiwan
Supermicro provides energy efficient server solutions optimized for GPU computing. Their portfolio includes 1U and 4U servers that support up to 10 GPUs, delivering the highest rack-level and node-level GPU density. Their new generation of solutions are optimized for machine learning applications using NVIDIA Pascal GPUs, with features like NVLink for high bandwidth GPU interconnect and direct low latency data access between GPUs. These solutions deliver the highest performance per watt for parallel workloads like machine learning training.
This document discusses NVIDIA's chips for automotive, HPC, and networking. For automotive, it describes the Tegra line of SOC chips used in cars like Tesla, and upcoming chips like Orin and Atlan. For HPC, it introduces the upcoming Grace CPU designed for giant AI models. For networking, it presents the BlueField line of data processing units (DPUs) including the new 400Gbps BlueField-3 chip and the DOCA software framework. The document emphasizes that NVIDIA's GPU, CPU, and DPU chips make yearly leaps while sharing a common architecture.
AMD Bridges the X86 and ARM Ecosystems for the Data Center AMD
Presentation by Lisa Su, senior vice president and general manager, Global Business Units, AMD regarding AMD’s announcement that it will design and build 64-bit ARM technology-based processors.
Lisa Spelman announced new Intel Xeon Scalable processors including Cascade Lake Advanced Performance processors with up to 48 cores, up to 2666 MHz DDR4 memory, and improved performance on HPC, AI and cloud workloads. Intel Software Guard Extensions provide advanced security capabilities to the new processors. Intel Optane persistent memory delivers affordable high capacity memory and high performance storage to support large scale analytics and database workloads.
AMD is introducing two new x86 CPU cores called "Bulldozer" and "Bobcat". Bulldozer is aimed at mainstream client and server markets, featuring improved performance and scalability. It uses a shared/dedicated resource approach to increase performance per watt. Bobcat is optimized for low power markets like cloud clients. It features a smaller, more efficient design to deliver high performance while minimizing power consumption and die size.
We’ve examined how we can rebuild inrastucture from scratch, but now let’s think outside the box, and inside the clouds. Before the zombie apocalypse began, many organizations were beginning to leverage public cloud infrastructures for a number of reasons.
The number of internet-connected devices is growing exponentially, enabling an increasing number of edge applications in environments such as smart cities, retail, and industry 4.0. These intelligent solutions often require processing large amounts of data, running models to enable image recognition, predictive analytics, autonomous systems, and more. Increasing system workloads and data processing capacity at the edge is essential to minimize latency, improve responsiveness, and reduce network traffic back to data centers. Purpose-built systems such as Supermicro’s short-depth, multi-node SuperEdge, powered by 3rd Gen Intel® Xeon® Scalable processors, increase compute and I/O density at the edge and enable businesses to further accelerate innovation.
Join this webinar to discover new insights in edge-to-cloud infrastructures and learn how Supermicro SuperEdge multi-node solutions leverage data center scale, performance, and efficiency for 5G, IoT, and Edge applications.
The document describes how the latest Intel® Advanced Vector Extensions 512 (Intel® AVX-512) instructions and Intel® Advanced Encryption Standard New Instructions (Intel® AES-NI) enabled in the latest Intel® 3rd Generation Xeon® Scalable Processor are used to significantly increase and achieve 1 Tb of IPsec throughput.
This document outlines various high performance computing (HPC) applications and technologies. It discusses computational areas like computational fluid dynamics, quantum mechanics, and climate simulation. It also mentions HPC systems from SGI including Altix, ICE, and UV architectures. These systems provide scalable shared memory and distributed memory computing utilizing technologies like NUMAlink interconnect and InfiniBand fabrics.
Hp cmu – easy to use cluster management utility @ hpcday 2012 kievVolodymyr Saviak
The document describes an overview presentation of the HP Insight Cluster Management Utility (CMU). It discusses the CMU's history and capabilities for provisioning, monitoring, and administering HPC clusters. Key points include that CMU can manage thousands of nodes, supports various Linux distributions, and provides tools for cloning, monitoring hardware and workloads, alerting and reactions to issues, and integrating partner software.
The document discusses the path to exascale computing and the challenges involved. It outlines existing trends like the end of Moore's law and major technology challenges. Technologies being developed to overcome issues include more efficient interconnects, memory, storage, and processors. The development of exascale systems will also require rethinking architectures and a paradigm shift in priorities like power efficiency. Significant code modernization efforts will be needed to effectively utilize exascale systems and harness massively parallel computing.
Jax 2013 - Big Data and Personalised MedicineGaurav Kaul
Global healthcare trends are driving an increase in data and a need for personalized medicine approaches. These include rising healthcare costs as populations age, with the average global age expected to rise from 10% to 21% over 60 by 2050. Next generation sequencing is driving down costs, enabling large amounts of genomic and other health data to be collected. This presents big data challenges to manage and analyze the data to enable personalized medicine approaches by 2020. Intel is working on solutions across the hardware and software stack to help with big healthcare data challenges including high performance computing, optimized software frameworks, data analytics methods, and use case examples like sequencing appliances.
Supermicro designed and implemented a rack-level cluster solution for San Diego Supercomputing Center (SDSC) optimized for their custom and experimental AI training and inferencing workloads, and meeting their environmental and TCO requirements. The project team will discuss the journey of designing and deploying our Rack Plug and Play cluster, and Shawn Strande, Dupty Director, SDSC, will be sharing his experience of partnering with the Supermicro team to solve his challgenges in HPC and AI.
The team will also share the technology that powers the SDSC Voyager Supercomputer, the Habana Gaudi AI system with 3rd Gen Intel® Xeon® Scalable processors for Deep Learning Training, and Habana Goya for Inferencing.
Watch the webinar: https://www.brighttalk.com/webcast/17278/517013
This document provides an overview of a new CPU capability called Intel® Speed Select
Technology – Base Frequency (Intel® SST-BF), which is available on select SKUs of 2nd
generation Intel® Xeon® Scalable processor (formerly codenamed Cascade Lake). The
document also includes benchmarking data and instructions on how to enable the
capability.
Value propositions of this capability include:
• Select SKUs of 2nd generation Intel® Xeon® Scalable processor (5218N, 6230N, and
6252N) offer a new capability called Intel® SST-BF.
• Intel® SST-BF allows the CPU to be deployed with an asymmetric core frequency
configuration.
• The placement of key workloads on higher frequency Intel® SST-BF enabled cores
can result in an overall system workload increase and potential overall energy
savings when compared to deploying the CPU with symmetric core frequencies
Fujitsu World Tour 2017 - Compute Platform For The Digital WorldFujitsu India
Fujitsu has decades of experience designing and manufacturing servers. Their PRIMERGY servers are known for best-in-class quality that ensures continuous operation with almost no unplanned downtimes. This is achieved through rigorous testing and manufacturing processes in their state-of-the-art factories in Germany. Fujitsu's demand-driven manufacturing approach allows them to produce servers flexibly based on current orders, enabling fast response times and fulfilling individual customer requests.
AMD and the new “Zen” High Performance x86 Core at Hot Chips 28AMD
The document summarizes a presentation about AMD's new "Zen" x86 CPU core architecture. The Zen architecture provides a 40% increase in instructions per clock compared to previous cores through improvements in the core engine, caches, floating point capabilities, and the addition of simultaneous multithreading. The Zen core was designed from the ground up to optimize performance and power efficiency across applications from notebooks to supercomputers.
dCUDA: Distributed GPU Computing with Hardware Overlapinside-BigData.com
Torsten Hoefler from ETH Zurich presented this deck at the Switzerland HPC Conference.
"Over the last decade, CUDA and the underlying GPU hardware architecture have continuously gained popularity in various high-performance computing application domains such as climate modeling, computational chemistry, or machine learning. Despite this popularity, we lack a single coherent programming model for GPU clusters. We therefore introduce the dCUDA programming model, which implements device-side remote memory access with target notification. To hide instruction pipeline latencies, CUDA programs over-decompose the problem and over-subscribe the device by running many more threads than there are hardware execution units. Whenever a thread stalls, the hardware scheduler immediately proceeds with the execution of another thread ready for execution. This latency-hiding technique is key to make best use of the available hardware resources. With dCUDA, we apply latency hiding at cluster scale to automatically overlap computation and communication. Our benchmarks demonstrate perfect overlap for memory bandwidth-bound tasks and good overlap for compute-bound tasks."
Watch the video presentation: http://wp.me/p3RLHQ-gCB
1. The document discusses Microsoft's data center strategy and optimization efforts to improve performance per watt, dollar, and reduce total cost of ownership.
2. Key aspects of Microsoft's strategy include modular pre-fabricated data center components, increased server and data center efficiency through workload analysis and low power processors, and reducing infrastructure costs through techniques like high-temperature cooling.
3. Microsoft has invested over $500 million to build large, efficient data centers like its Chicago facility, and is working towards new designs with no mechanical cooling and ultra-modularity to further reduce costs and increase scalability.
Microsoft Project Olympus AI Accelerator Chassis (HGX-1)inside-BigData.com
In this video from the Open Compute Summit, Siamak Tavallaei from Microsoft presents an overview of the Microsoft Project Olympus AI Accelerator Chassis, also known as the HGX-1.
Watch the presentation video: http://wp.me/p3RLHQ-guX
Sign up for our insideHPC Newsletter: http://insidehpc.com/newsletter
The GIST AI-X Computing Cluster provides powerful accelerated computation resources for machine learning using GPUs and other hardware. It includes DGX A100 and DGX-1V nodes with 8 NVIDIA A100 or V100 GPUs each, connected by high-speed networking. The cluster uses Singularity containers, Slurm scheduling, and Ceph storage. It allows researchers to request resources, build container images, and run distributed deep learning jobs across multiple GPUs.
Evolution of Supermicro GPU Server SolutionNVIDIA Taiwan
Supermicro provides energy efficient server solutions optimized for GPU computing. Their portfolio includes 1U and 4U servers that support up to 10 GPUs, delivering the highest rack-level and node-level GPU density. Their new generation of solutions are optimized for machine learning applications using NVIDIA Pascal GPUs, with features like NVLink for high bandwidth GPU interconnect and direct low latency data access between GPUs. These solutions deliver the highest performance per watt for parallel workloads like machine learning training.
This document discusses NVIDIA's chips for automotive, HPC, and networking. For automotive, it describes the Tegra line of SOC chips used in cars like Tesla, and upcoming chips like Orin and Atlan. For HPC, it introduces the upcoming Grace CPU designed for giant AI models. For networking, it presents the BlueField line of data processing units (DPUs) including the new 400Gbps BlueField-3 chip and the DOCA software framework. The document emphasizes that NVIDIA's GPU, CPU, and DPU chips make yearly leaps while sharing a common architecture.
AMD Bridges the X86 and ARM Ecosystems for the Data Center AMD
Presentation by Lisa Su, senior vice president and general manager, Global Business Units, AMD regarding AMD’s announcement that it will design and build 64-bit ARM technology-based processors.
Lisa Spelman announced new Intel Xeon Scalable processors including Cascade Lake Advanced Performance processors with up to 48 cores, up to 2666 MHz DDR4 memory, and improved performance on HPC, AI and cloud workloads. Intel Software Guard Extensions provide advanced security capabilities to the new processors. Intel Optane persistent memory delivers affordable high capacity memory and high performance storage to support large scale analytics and database workloads.
AMD is introducing two new x86 CPU cores called "Bulldozer" and "Bobcat". Bulldozer is aimed at mainstream client and server markets, featuring improved performance and scalability. It uses a shared/dedicated resource approach to increase performance per watt. Bobcat is optimized for low power markets like cloud clients. It features a smaller, more efficient design to deliver high performance while minimizing power consumption and die size.
We’ve examined how we can rebuild inrastucture from scratch, but now let’s think outside the box, and inside the clouds. Before the zombie apocalypse began, many organizations were beginning to leverage public cloud infrastructures for a number of reasons.
The number of internet-connected devices is growing exponentially, enabling an increasing number of edge applications in environments such as smart cities, retail, and industry 4.0. These intelligent solutions often require processing large amounts of data, running models to enable image recognition, predictive analytics, autonomous systems, and more. Increasing system workloads and data processing capacity at the edge is essential to minimize latency, improve responsiveness, and reduce network traffic back to data centers. Purpose-built systems such as Supermicro’s short-depth, multi-node SuperEdge, powered by 3rd Gen Intel® Xeon® Scalable processors, increase compute and I/O density at the edge and enable businesses to further accelerate innovation.
Join this webinar to discover new insights in edge-to-cloud infrastructures and learn how Supermicro SuperEdge multi-node solutions leverage data center scale, performance, and efficiency for 5G, IoT, and Edge applications.
The document describes how the latest Intel® Advanced Vector Extensions 512 (Intel® AVX-512) instructions and Intel® Advanced Encryption Standard New Instructions (Intel® AES-NI) enabled in the latest Intel® 3rd Generation Xeon® Scalable Processor are used to significantly increase and achieve 1 Tb of IPsec throughput.
This document outlines various high performance computing (HPC) applications and technologies. It discusses computational areas like computational fluid dynamics, quantum mechanics, and climate simulation. It also mentions HPC systems from SGI including Altix, ICE, and UV architectures. These systems provide scalable shared memory and distributed memory computing utilizing technologies like NUMAlink interconnect and InfiniBand fabrics.
Hp cmu – easy to use cluster management utility @ hpcday 2012 kievVolodymyr Saviak
The document describes an overview presentation of the HP Insight Cluster Management Utility (CMU). It discusses the CMU's history and capabilities for provisioning, monitoring, and administering HPC clusters. Key points include that CMU can manage thousands of nodes, supports various Linux distributions, and provides tools for cloning, monitoring hardware and workloads, alerting and reactions to issues, and integrating partner software.
GPU HPC Clusters document discusses GPU cluster research at NCSA including early GPU clusters like QP and Lincoln, follow-up clusters like AC that expanded GPU resources, and eco-friendly cluster EcoG. It describes ISL research in GPU and heterogeneous computing including systems software, runtimes, tools and application development.
Graph500 and Green Graph500 benchmarks on SGI UV2000 @ SGI UG SC14Yuichiro Yasui
The document discusses Graph500 and Green Graph500 benchmarks for evaluating graph processing performance on the SGI UV2000 system. It provides an overview of the benchmarks and describes testing various graph workloads, including social networks and road networks, on different hardware from smartphones to supercomputers. The authors aim to optimize breadth-first search (BFS) graph algorithms on the NUMA-based SGI UV2000 without using MPI through NUMA-aware techniques.
The document provides an overview of Project Ultraviolet, SGI's next-generation high performance computing architecture. It describes UV's scalable ccNUMA architecture using Intel Xeon processors with SGI's NUMAlink interconnect. UV can scale to 2048 cores and 16TB of memory in a single system. It offers accelerated performance through the MPI Offload Engine and massive memory-mapped I/O. UV also aims for industry-leading rack-level power efficiency of 80% and low total cost of ownership through standard x86 components.
NUMA-aware Scalable Graph Traversal on SGI UV SystemsYuichiro Yasui
The document discusses NUMA-aware scalable graph traversal on SGI UV systems. It proposes an efficient NUMA-aware breadth-first search (BFS) algorithm for large-scale graph processing by pruning remote edge traversals. Numerical results on SGI UV 300 systems with 32 sockets show the algorithm achieves 219 billion traversed edges per second (GTEPS), setting a new single-node performance record on the Graph500 benchmark.
NUMA-optimized Parallel Breadth-first Search on Multicore Single-node SystemYuichiro Yasui
The document proposes a NUMA-optimized parallel breadth-first search (BFS) algorithm for multicore systems. It discusses how the hybrid BFS algorithm combines top-down and bottom-up approaches but can result in unnecessary edge traversals. The proposal distributes the graph columns to each NUMA node's local memory and binds threads and data to improve locality. It uses a library called ULIBC to intelligently manage CPU affinity and NUMA considerations. Numerical results show the NUMA-optimized hybrid BFS achieves up to 2.2x speedup over the original algorithm.
Altreonic was spun off in 2008 from Eonic Systems to focus on real-time operating systems using formal techniques. Their OpenComRTOS is a small, network-centric real-time OS that uses CSP concurrency and can scale from 1 to over 10,000 nodes. It provides priority-based communication and fault tolerance and has been implemented on many heterogeneous platforms from DSPs to many-core systems.
The Monash Sun Grid Programme provides high-performance and high-throughput computing resources at Monash University. It consists of a central high-performance compute cluster called the Monash Sun Grid with over 1,650 CPU cores and over 5.7TB of RAM. The cluster hardware has been upgraded over time with newer node types providing more cores, memory, and faster processing. Specialist support is available to help users with tasks like software installation, job preparation, and performance optimization. Usage of the cluster has grown significantly in recent years in terms of CPU hours consumed and number of active users. Looking ahead, the programme aims to continue hardware refreshes, add more computing nodes, improve storage infrastructure, and better integrate with grid middleware
This document discusses how data is increasingly dominating high performance computing workloads. It notes that while computing power doubles every two years, data storage and movement capabilities are not keeping pace. This is leading to a "data tsunami" as experiments and simulations generate terabytes of data per day. The document then summarizes Sun Microsystems' end-to-end infrastructure for data-centric HPC workflows, including their Lustre parallel storage system, unified storage, tape archives, high performance computing blades, and InfiniBand switches. It positions Sun as uniquely able to deliver an integrated solution from computation to long-term data retention to help users cope with the challenges posed by rapidly growing datasets.
The document discusses how Pure Storage's FlashBlade storage system is designed to be a data hub that can power various modern data and analytics workloads including AI, machine learning, data warehousing, and streaming analytics. It provides high throughput, scale-out performance for file and object storage, and is purpose built to deliver the performance needed for these next generation workloads. FlashBlade uses a scale-out architecture with blades, networking, and software that allows for linear scaling of performance and capacity.
[RakutenTechConf2013] [A-3] TSUBAME2.5 to 3.0 and Convergence with Extreme Bi...Rakuten Group, Inc.
Rakuten Technology Conference 2013
"TSUBAME2.5 to 3.0 and Convergence with Extreme Big Data"
Satoshi Matsuoka
Professor
Global Scientific Information and Computing (GSIC) Center
Tokyo Institute of Technology
Fellow, Association for Computing Machinery (ACM)
How to Terminate the GLIF by Building a Campus Big Data Freeway SystemLarry Smarr
12.10.11
Keynote Lecture
12th Annual Global LambdaGrid Workshop
Title: How to Terminate the GLIF by Building a Campus Big Data Freeway System
Chicago, IL
Container Attached Storage (CAS) with OpenEBS - Berlin Kubernetes Meetup - Ma...OpenEBS
The OpenEBS project has taken a different approach to storage when it comes to containers. Instead of using existing storage systems and making them work with containers; what if you were to redesign something from scratch using the same paradigms used in the container world? This resulted in the effort of containerizing the storage controller. Also, as applications that consume storage are changing over, do we need a scale-out distributed storage systems?
Riken's Fujitsu K computer is the world's fastest supercomputer, with a peak performance of over 11 petaflops. It uses a homogeneous architecture of over 700,000 SPARC64 VIIIfx processors connected via a high-speed interconnect. Looking ahead, future exascale supercomputers in the 2018 timeframe are projected to have over 1 exaflop of peak performance, use over 1 billion processing cores, and consume around 20 megawatts of power. Significant technological advancements will be required across hardware and software to achieve exascale capabilities.
The document discusses Oracle's Exalogic engineered systems. It begins with an overview of Oracle's engineered systems strategy and then focuses on Exalogic. Key points about Exalogic include:
- It features optimized hardware including servers, an InfiniBand fabric, and integrated storage.
- Exalogic can be deployed in quarter rack, half rack, full rack, or multi-rack configurations allowing for scaling.
- The document reviews Exalogic's virtualization capabilities and how it supports application consolidation and multi-tenancy.
- High availability, redundancy, and seamless scalability are emphasized as benefits of Exalogic.
In this deck from the HPC User Forum in Santa Fe, Peter Hopton from Iceotope presents: European Exascale System Interconnect & Storage.
"A new Exascale computing architecture using ARM processors is being developed by a European consortium of hardware and software providers, research centers, and industry partners. Funded by the European Union’s Horizon2020 research program, a full prototype of the new system is expected to be ready by 2018."
The project, called ExaNeSt, is based on ARM processors, originally developed for mobile and embedded applications, similar to another EU project, Mont Blanc, which also aims to design a supercomputer architecture using an ARM based supercomputer. Where ExaNeSt differs from Mont Blanc, however, is a focus on networking and on the design of applications. ExaNeSt is co-designing the hardware and software, enabling the prototype to run real-life evaluations – facilitating a stable, scalable platform that will be used to encourage the development of HPC applications for use on this ARM based supercomputing architecture.
Watch the video:
Learn more: http://www.iceotope.com/
and
http://hpcuserforum.com
Sign up for our insideHPC Newsletter: http://insidehpc.com/newsletter
The Roadrunner supercomputer is housed at Los Alamos National Laboratory and was the second fastest in the world at 1.7 petaflops. It uses a hybrid architecture of AMD and IBM processors to simulate nuclear materials and ensure safety of the US nuclear arsenal. Roadrunner occupies 6000 square feet and consists of 18 connected units each with 180 blade servers containing both Opteron and PowerXCell processors for a total of over 100,000 cores.
El Barcelona Supercomputing Center (BSC) fue establecido en 2005 y alberga el MareNostrum, uno de los superordenadores más potentes de España. Somos el centro pionero de la supercomputación en España. Nuestra especialidad es la computación de altas prestaciones - también conocida como HPC o High Performance Computing- y nuestra misión es doble: ofrecer infraestructuras y servicio de supercomputación a los científicos españoles y europeos, y generar conocimiento y tecnología para transferirlos a la sociedad. Somos Centro de Excelencia Severo Ochoa, miembros de primer nivel de la infraestructura de investigación europea PRACE (Partnership for Advanced Computing in Europe), y gestionamos la Red Española de Supercomputación (RES). Como centro de investigación, contamos con más de 456 expertos de 45 países, organizados en cuatro grandes áreas de investigación: Ciencias de la computación, Ciencias de la vida, Ciencias de la tierra y aplicaciones computacionales en ciencia e ingeniería.
Open stack in action cern _openstack_accelerating_scienceeNovance
This document discusses CERN's adoption of OpenStack for its computing infrastructure. CERN operates the Large Hadron Collider and needs to manage thousands of servers to process massive amounts of experimental data. It is moving to OpenStack to improve operational efficiency, resource utilization, and responsiveness. The current status is a pre-production OpenStack deployment using Nova, Hyper-V, and other components. CERN contributes code and expertise back to the OpenStack community.
20121205 open stack_accelerating_science_v3Tim Bell
CERN is a large particle physics laboratory located between Geneva and France that is seeking answers to fundamental questions about the universe. It operates the Large Hadron Collider (LHC) and collects massive amounts of data that are stored and analyzed using a worldwide computing grid. CERN is adopting open source tools like OpenStack for its computing infrastructure to better manage scaling its systems with limited staff increases. It is contributing to open source communities and training staff on these tools to gain valuable skills applicable outside of CERN.
High Performance Cyberinfrastructure Enabling Data-Driven Science in the Biom...Larry Smarr
11.04.06
Joint Presentation
UCSD School of Medicine Research Council
Larry Smarr, Calit2 & Phil Papadopoulos, SDSC/Calit2
Title: High Performance Cyberinfrastructure Enabling Data-Driven Science in the Biomedical Sciences
The document discusses the evolution of computer architectures from early technological achievements like the transistor and integrated circuit. It describes increasing transistor densities following Moore's Law. Future technologies will focus on increasing core counts while decreasing cycle times and voltages. Performance will come from parallelism rather than clock speed increases due to heat limitations. The document outlines challenges in scaling to exascale systems by 2018.
The document discusses CSC's high performance computing capabilities and developments from 2012-2014. It summarizes that CSC upgraded its Sisu and Taito systems in 2014 with new Intel Haswell CPUs, increasing cores by 50% and reducing energy usage. The upgrades boosted Sisu's performance to 1700 TFlops, making it the 37th most powerful system in the world. CSC now provides a total of 2.54 PFlops of computing power and is the most powerful academic computing facility in the Nordic countries.
Exalogic is an impressive piece of hardware offering immense performance. However the smallest configuration is 96 cores, 768 GB memory and a 40TB SAN... way bigger than many smaller customers could imagine using (even including test environments).
This session takes a look at how you could use modern server technology, such as blades, to build a smaller version of Exalogic, and yet still benefit from some of the cost savings from sophisticated automation. This will include a case study of a mid-sized installation where these techniques have been used.
Delivered on 5 December 2011 at UKOUG 2011 by Simon Haslam.
Altair - compute manager your gateway to hpc cloud computing with pbs profess...Volodymyr Saviak
Altair's Compute Manager is a cloud computing solution that allows users to submit, monitor, and manage HPC jobs on distributed resources powered by PBS Professional. It provides a web interface for easily configuring and running simulations, visualizing results, and managing compute jobs without requiring advanced computer skills. The system leverages Altair's HyperWorks platform and can be customized for various engineering domains.
The document discusses new advancements in high-performance computing (HPC) interconnect technology from Mellanox Technologies. It outlines how Mellanox's FDR InfiniBand has become the most commonly used interconnect solution for HPC, connecting more of the world's fastest supercomputers. It also presents Mellanox's roadmap and new products that support higher speeds and capabilities to pave the way for exascale computing through solutions like Connect-IB and optimizations for GPUs and accelerators.
The document discusses best practices for accelerating innovation with high-performance computing (HPC). It notes that HPC is used across many industries like engineering, manufacturing, life sciences, geosciences, government, academia, finance, media, and entertainment. It also discusses having a flexible HPC infrastructure that can adapt to changing needs.
The document discusses Mellanox's interconnect solutions for high performance computing (HPC). It describes Mellanox's end-to-end connectivity portfolio including host fabric adapters, switches, cables and software. It highlights recent introductions of 56Gb/s FDR InfiniBand technology and hardware acceleration solutions for collectives and GPUs. Mellanox's solutions are powering several top supercomputers and InfiniBand has become the dominant interconnect for HPC systems.
Building RAG with self-deployed Milvus vector database and Snowpark Container...Zilliz
This talk will give hands-on advice on building RAG applications with an open-source Milvus database deployed as a docker container. We will also introduce the integration of Milvus with Snowpark Container Services.
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024Albert Hoitingh
In this session I delve into the encryption technology used in Microsoft 365 and Microsoft Purview. Including the concepts of Customer Key and Double Key Encryption.
Sudheer Mechineni, Head of Application Frameworks, Standard Chartered Bank
Discover how Standard Chartered Bank harnessed the power of Neo4j to transform complex data access challenges into a dynamic, scalable graph database solution. This keynote will cover their journey from initial adoption to deploying a fully automated, enterprise-grade causal cluster, highlighting key strategies for modelling organisational changes and ensuring robust disaster recovery. Learn how these innovations have not only enhanced Standard Chartered Bank’s data infrastructure but also positioned them as pioneers in the banking sector’s adoption of graph technology.
How to Get CNIC Information System with Paksim Ga.pptxdanishmna97
Pakdata Cf is a groundbreaking system designed to streamline and facilitate access to CNIC information. This innovative platform leverages advanced technology to provide users with efficient and secure access to their CNIC details.
TrustArc Webinar - 2024 Global Privacy SurveyTrustArc
How does your privacy program stack up against your peers? What challenges are privacy teams tackling and prioritizing in 2024?
In the fifth annual Global Privacy Benchmarks Survey, we asked over 1,800 global privacy professionals and business executives to share their perspectives on the current state of privacy inside and outside of their organizations. This year’s report focused on emerging areas of importance for privacy and compliance professionals, including considerations and implications of Artificial Intelligence (AI) technologies, building brand trust, and different approaches for achieving higher privacy competence scores.
See how organizational priorities and strategic approaches to data security and privacy are evolving around the globe.
This webinar will review:
- The top 10 privacy insights from the fifth annual Global Privacy Benchmarks Survey
- The top challenges for privacy leaders, practitioners, and organizations in 2024
- Key themes to consider in developing and maintaining your privacy program
Introducing Milvus Lite: Easy-to-Install, Easy-to-Use vector database for you...Zilliz
Join us to introduce Milvus Lite, a vector database that can run on notebooks and laptops, share the same API with Milvus, and integrate with every popular GenAI framework. This webinar is perfect for developers seeking easy-to-use, well-integrated vector databases for their GenAI apps.
Let's Integrate MuleSoft RPA, COMPOSER, APM with AWS IDP along with Slackshyamraj55
Discover the seamless integration of RPA (Robotic Process Automation), COMPOSER, and APM with AWS IDP enhanced with Slack notifications. Explore how these technologies converge to streamline workflows, optimize performance, and ensure secure access, all while leveraging the power of AWS IDP and real-time communication via Slack notifications.
Climate Impact of Software Testing at Nordic Testing DaysKari Kakkonen
My slides at Nordic Testing Days 6.6.2024
Climate impact / sustainability of software testing discussed on the talk. ICT and testing must carry their part of global responsibility to help with the climat warming. We can minimize the carbon footprint but we can also have a carbon handprint, a positive impact on the climate. Quality characteristics can be added with sustainability, and then measured continuously. Test environments can be used less, and in smaller scale and on demand. Test techniques can be used in optimizing or minimizing number of tests. Test automation can be used to speed up testing.
Dr. Sean Tan, Head of Data Science, Changi Airport Group
Discover how Changi Airport Group (CAG) leverages graph technologies and generative AI to revolutionize their search capabilities. This session delves into the unique search needs of CAG’s diverse passengers and customers, showcasing how graph data structures enhance the accuracy and relevance of AI-generated search results, mitigating the risk of “hallucinations” and improving the overall customer journey.
UiPath Test Automation using UiPath Test Suite series, part 5DianaGray10
Welcome to UiPath Test Automation using UiPath Test Suite series part 5. In this session, we will cover CI/CD with devops.
Topics covered:
CI/CD with in UiPath
End-to-end overview of CI/CD pipeline with Azure devops
Speaker:
Lyndsey Byblow, Test Suite Sales Engineer @ UiPath, Inc.
Communications Mining Series - Zero to Hero - Session 1DianaGray10
This session provides introduction to UiPath Communication Mining, importance and platform overview. You will acquire a good understand of the phases in Communication Mining as we go over the platform with you. Topics covered:
• Communication Mining Overview
• Why is it important?
• How can it help today’s business and the benefits
• Phases in Communication Mining
• Demo on Platform overview
• Q/A
Maruthi Prithivirajan, Head of ASEAN & IN Solution Architecture, Neo4j
Get an inside look at the latest Neo4j innovations that enable relationship-driven intelligence at scale. Learn more about the newest cloud integrations and product enhancements that make Neo4j an essential choice for developers building apps with interconnected data and generative AI.
1. Welcome to the outer limits
HPC Day, Kiev October 14 2011
Martin Jalový, Systems Engineer SGI
2. Our Differentiated View of the Market
A Trusted Leader in
Technical Computing
“… As we enter the next generation, moving from petaflops to
exaflops, SGI has emerged a trusted leader in technical
computing. Supporting some of the industry's most mission critical
and large-scale computing applications.”
-Earl Joseph, program vice president, high-performance systems
Company Confidential July 2011
3. Škoda Auto – ICE 8200, ICE 8400 Q3 2011
????? Processor Cores, ICE (was 3328c in 2010), UV
??? Tflop/s Peak Performance
???? TB Memory
?,? PB storage, Nexis 9000, ISS3500 fast scratch, CXFS
specialized nodes (eg. prepost n., each 144GB mem.
Grid engine
Company Confidential July 2011
4. NIIFI, University Hungaria – ICE, UV 1000 Q2 2011
Professor Stephen
Hawking inspecting
the final Altix UV
installation
1536 Cores ICE 8400
1152 Cores UV 1000
20,2+12,2 Tflop/s Peak Performance as Planck
Strategic projects, such
6,1 +6,1 TB Memory
satellite analysis codes
Storage, Lustre
Water cooled
Large simulation requirements
of theoretical ideas and
comparison with vast new
oбразовать observational data sets
Company Confidential July 2011
5. PCSS – Poznan Univ., GPU cluster 227n, UV 1000 Q4 2011
227 nodes with mix 1 or 2 GPU per node
7264 Processor Cores (AMD Interlagos)a
~50 Tflop/s Peak Performance
Water cooled
2048 Processor Cores, UV 1000
21,7 Tflop/s Peak Performance
16 TB Memory
Water cooled
Company Confidential July 2011
6. COSMOS, Cambridge University - UV 1000
August 2010
768 Processor Cores UV 1000
63 Tflop/s Peak Performance
2 TB Shared Memory
64TB IS4100 storage, CXFS Professor Stephen
MOAB Grid suite Hawking inspecting
Water cooled the final Altix UV
installation
Strategic projects, such as Planck
satellite analysis codes
Large simulation requirements
of theoretical ideas and
comparison with vast new
observational data sets
Company Confidential July 2011
7. Los Alamos National Lab – Altix XE clusters
Conejo project 51,4Tfl, Mapache project 49,2Tfl
10096 Processor Cores, Altix XE Peralta, Coyote
101,6 Tflop/s Peak Performance
30,2 TB Memory
Climate, Ocean, Advanced Simulation and
Company Confidential July 2011 National Nuclear Security Administration
8. LRZ - Altix 4700 (Munich)
Largest globally shared memory system in the world
9728 Processor Cores
63 Tflop/s Peak Performance
NEW!
39 TB Shared Memory
2048 Processor Cores, UV 1000
660TB RAID Storage, CXFS
Company Confidential July 2011 40GB/s
9. Total – bigest commercial system in the world
234 TFlops CPU
Expansion 2011
7680 cores
Westmere
73 TB Mem
2,5 PB Lustre
40GB/s
Company Confidential July 2011
10. RossHydroMet
Moscow, St. Petersburg, Novosibirsk, Khabarovsk
Moscow center
1664 cores Altix 4700, 11 TFlops peak, 6.6TB memory (4GB/core)
1416 cores Altix ICE, 2.5 TB (2GB/core)
other servers 12x Altix 450, 17x Altix XE,
CXFS, DMF, 60TB online, 200TB nearline, 86TB tape library
Company Confidential July 2011
11. Altix® ICE:
CINES - Grand Equipement National de Calcul Intensif (GENCI)
• System installed at
CINES, France's
National Computer
Center for Higher
Education; supports
French researchers
23040 cores Altix ICE in variety of
267 Tflops Peak Performance research areas, and
103 TB of Memory connect to the
0,8 PB storage (IS 4600), DMF, Lustre RENATER French
high-speed network
20GB/s IO performance
and to European
Community
Slide 11
Company Confidential July 2011 infrastructure.
12. HLRN – HLRN Link – 2 sites = 1 system
One Grid – One Workload – One Filesystem
Norddeutscher Verbund für Hoch- und Höchstleistungsrechnen
Single userbase/accouting
Single workload managed
across two sites
MPP1 MPP2
MPP1 MPP2
HLRN-Link Infiniband
SMP1 SMP2
Infiniband
SMP1 SMP2
30976 cores (ICE 20480c+UV1000 9728c)
339 Tflops Peak Performance,
114 TB of Memory
1.15 PB Storage, Lustre, LSI, target 2.3 PB
2x18GB/s IO bandwidth
Company Confidential July 2011
13. NASA Pleiades Supercomputer
number 7 in Top500 06/2011
112540 cores Altix ICE, 184 racks
1,009 PFlops RPeak Performance
188 TB of Memory (1GB/core)
storage SGI Nexis 9000, 7 Lustre cluster,
12 DDN RAIDS 6.9 PB
Company Confidential July 2011 two IB network 11D hypercube topology
15. Altix® XE: An Enterprise Solution to Keep Paris
Taxis Moving for Les Taxis Bleus
The Challenge The Solution
An increasing burden on the
• SGI® Altix® XE server with 8GB of memory running
existing infrastructure had made
it impossible to guarantee 100 Red Hat®Enterprise Linux® v.4
percent availability of the • Connects to an Oracle 10g database and 1TB of SGI
reservation management system, InfiniteStorage 220 direct attached Fiber Channel
resulting in potential lost • Services 120 dispatch agents, and drives Taxis
business.
Bleus’ internal accounting system.
• Redundant configuration backs up the primary
system for no-compromise reliability, availability
and serviceability (RAS)
“Now back-ups happen faster, and we don’t have to
choose between backing up our enterprise data to tape or actually
managing our bookings and billings.”
Patrick Del’Vecchio, Director of IT Systems, Les Taxis Bleus
Slide 15
Company Confidential July 2011
16. Our main Markets
CLOUD
CLOUD STORAGE
BIG
HPC Web 2.0
SaaS
Active Archive
Storage Servers DATA
HP-HD Raid
Commercial Hadoop
Scientific Databases
Warehouses
Company Confidential July 2011
17. SGI Solutions Across Applications
Oil & Gas HPC Cluster Large Trading Platforms Large Cloud Provider
M45 ICE Cube HPC National Lab Video Streaming
Company Confidential July 2011
18. SGI® Technology @ Work
DEFENSE & INTELLIGENCE DIGITAL MEDIA ENERGY FINANCIAL SERVICES
CLOUD MANUFACTURING RESEARCH & EDUCATION …AND MORE
Company Confidential July 2011
20. SGI OpenFOAM® Ready for Cyclone
Customer : iVEC and Curtin
Technical Applications Portal University Australia
Powered by
User Problem: Solving large scale CFD
Su
bm problems like simulating wind flows
its
J ob in the capital city of Perth.
Solution: OpenFOAM scaled on SGI
Cyclone better (1024 cores) and
was 20x faster than on Amazon
EC2.
Source: Dr Andrew King, Department of Mechanical Engineering Curtin, University of Technology, Australia
Company Confidential July 2011
21. SGI: Servers, Storage, Experience
Would you attach
caravan
to the F1?
Company Confidential July 2011
22. Datacenter Technologies : Eco-Logical™
DCE = PUE-1 = equip x facility-1
SGI top in its class
Savings from
direct water World Largest
cooling doors. datacenters,
hundreds installed design
experience
Savings from
12VDC rack
distribution
(480VDC
datacenter*) World Leader:
mobile,
containerized
datacenter
20’/40’ units
Savings from
90+% efficient
power supplies.
80 PLUS awards
* evaluating
Company Confidential July 2011
23. Power Supplies : High Efficiency
• AC power supplies up
to 92.5% efficiency*
• 80Plus validated at 91.2%
typical efficiency 90.6% SGI
Power Supply
• 48VDC power supplies
up to 96.5% efficiency*
• High-efficiency
redundant cabinet-level
48VDC in Rackable DC
SGI 450W AC Power Supply Efficiency
cabinets and PowerXE
(Validated by 80Plus.org)
distribution in
CloudRack C2
Company Confidential July 2011 * Peak efficiency based on power supply specification
25. Server Products
ICE Cube® Air Modular Datacenter
The
Power of Scale-Out Scale-Up
One
CloudRack™ X2 Rackable™ & Altix® ICE SGI Prism™ XL Altix UV
Octane® III CloudRack™ C2 Blade Architecture Petaflop in a Rack Shared-Memory Systems
Origin® 400 1U, 2U, 3U, 4U HPC Optimized Optimized
Design-to-Order Accelerator
Leader
Company Confidential July 2011
28. Customer Challenges
Need Faster Simulation on Large Datasets
– Altix: Use Fastest CPU’s available
Need to Simulate Very Large Datasets
– Altix: Extremely Large Memory Systems
Fast Delivery to Production and Results
– Altix: Standard Operating Systems
Ability to Scale System as Requirements Increase
– Altix: Range of Products based on same architecture
Company Confidential July 2011
29. FY11 was Good
TECHNICAL UNIVERSITY OF
DENMARK DTU
Company Confidential July 2011
31. Altix® UV
World’s Fastest Supercomputer
– World Record SPECint_rate, SPECfp_rate, SPECjbb2005 Performance
– High speed NUMAlink® 5 interconnect (15 GB/sec)
– MPI offload engines maximize efficiency
– Direct access to global data sets up to 16TB
– Compelling performance regardless of type of application
Scalable
– Single system image scales up to 2048 cores & 16TB memory
– Investment protection: start with four sockets, scale up over time
Open Platform
– Leverages Intel® Xeon® 7600 (“Westmere-EX”) processors
– Runs industry-standard x86 operating systems & application code
Company Confidential July 2011
32. SGI Altix UV and Windows Server
2008 R2
The world’s most powerful x86 Windows® Server
platform: the SGI® Altix® UV, world’s fastest and most
scalable supercomputer
– Now certified up to 256 cores, 2 TB shared memory
– The only platform certified up to the full processor AND
memory limits of Windows Server 2008 R2
Industry standard Intel® Xeon® processors
Single instance of Windows Server
Runs standard Microsoft software:
– Windows Server 2008 R2 Datacenter Edition
– Unlimited virtualization with Hyper-V
– Microsoft SQL Server
Expandability: SGI UV enables new releases of
SGI Altix UV Microsoft software to scale to their limits
Company Confidential July 2011
33. Altix UV: Supporting the Microsoft
Platform
Microsoft Stack Big Data Technical Apps
Data Warehousing
Company Confidential July 2011
34. Customer benefits
Altix UV Delivers
– Higher performance and better price performance –
delivers the capabilities of SGI Altix UV to
Windows Server users
Customer Benefits
– Faster data access for a range of data intensive computing
(more data in memory)
– Larger databases without partitioning
– Faster computing using lower latency
– Easier management
– Consolidated resources
– Quicker and more complete business decisions
SGI Altix UV
Company Confidential July 2011
35. SGI Altix UV running Microsoft Windows
Server 2008 R2
World’s largest Windows platform:
256 cores, 2 TB memory
4x the number of cores over HP*
2.6x the number of cores of IBM
Industry defining application scalability for
Windows applications: 64 Intel Xeon Physical Cores
Compute intensive
Data intensive
SGI Altix UV *HP ProLiant
DL 980 G7
Company Confidential July 2011
38. Reliability Altix ICE 8200, Altix ICE 8400
Cable-free, Redundant Components
Cable-free blade
enclosures: 128 cores
and no cables
Redundant, hot swap
power and cooling
Fully Buffered DIMMS
to reduce transient
errors
Blade design provides
rapid serviceability
InfiniBand backplane
To THIS
From this (traditional for high signal
racked cluster)… reliability
Company Confidential July 2011
39. SGI® Altix® ICE 8400
Designed for High-Performance Computing
Performance Density: Up to 1536 Cores and
14.13 TFlops per Rack / 8.3 ft2 (0.77 m2)
SGI® Altix® ICE Compute Blade
Up to 12-Core, 96GB, 2-IB
SGI® Altix® ICE Compute Blade
Up to 24-Core, 128GB, 2-IB
Altix ICE Rack:
• 42U rack (30” W x 40” D)
• 4 Cable-free blade enclosures, each with up to 16 2-Socket nodes
• Up to 128 DP Intel® Xeon® or AMD Opteron™ 6100 sockets
• Single-plane or Dual-plane IB QDR interconnect
• Minimal switch topology simplifies scaling to 1000s of nodes
Slide 39
Company Confidential July 2011
40. Introducing Altix® ICE 8400
No-compromise, high performance
blade design for scale-out clusters
Tightly integrated
Greatly increased flexibility:
– Five different compute blades
– Choice of processor type, network
topology, system fabric and storage
– Value and performance-optimized
configurations
QDR InfiniBand- through and through
Company Confidential July 2011
41. Flexible Compute Blade Options
Intel® Xeon® 5500/5600 or AMD Opteron™ 6100
processors
Intel blades feature 12 DIMM slots and up to 768
cores/cabinet. Up to 130W processors are supported.
AMD blades feature 16 DIMM slots and up to 1536
cores/cabinet* Up to 105W processors are supported
Choice of three on-board Mellanox® ConnectX-2
InfiniBand HCA configurations
– Single-port, dual-port or two single-port chipset(s)
Option for 2.5” storage on the node (SSD and/or HDD)
Intel Intel
NHM/ WM-EP NHM/ WM-EP
Intel Intel Intel Intel Processor Processor
NHM/ WM-EP NHM/ WM-EP NHM/ WM-EP NHM/ WM-EP
Processor Processor Processor Processor
Single
Tylersburg Tylersburg IB4x QDR Tylersburg
HCA
Single x4 Chipset Dual x4 Chipset Chipset
Backplane Connector
Backplane Connector
Backplane Connector
QDR IB QDR IB
HCA HCA Single
IB4x QDR
HCA
ICH10 ICH10
PHY PHY
ICH10
BMC BMC PHY
PHY PHY BMC
PHY
Flash Memory SIO FWH Flash Memory SIO FWH
Flash Memory SIO FWH
Company Confidential July 2011 * AMD Opteron blades planned for Q3CY10.
42. Flexibility in Networking Topologies
Hypercube Topology:
- Lowest network infrastructure cost
- Well suited for "nearest neighbor" type MPI
communication patterns
Enhanced Hypercube Topology:
- Increased bisectional bandwidth per node at only
a small increase in cost
- Well suited for larger node count MPI jobs
All-to-All Topology:
- Maximum bandwidth at lowest latency for up to
128 nodes
- Well suited for "all-to-all" MPI communication
patterns.
Robust integrated switch
blade design enables industry- Fat Tree Topology:
leading bisectional bandwidth - Highest network infrastructure cost. Requires
external switches.
at ultra-low latency!
- Well suited for "all-to-all" type MPI
communication patterns
Company Confidential July 2011
43. SGI® Altix® ICE 8400 IRU Topology –
Hypercube optimized
Compute 03
Compute 03
Compute 03
Compute 03
Compute 07
Compute 05
Compute 05
Compute 05
Compute 05
Compute 04
Compute 04
Compute 04
Compute 04
Compute 02
Compute 01
Compute 00
Compute 06
Plane-1 Plane-2
Switch Blade
Switch Blade
36-Port QDR
36-Port QDR
InfiniBand
InfiniBand
1st Hypercube Dimension is
Quad-Linked on Backplane
Switch Blade
Switch Blade
36-Port QDR
36-Port QDR
InfiniBand
InfiniBand
Compute 15
Compute 13
Compute 13
Compute 13
Compute 13
Compute 11
Compute 11
Compute 11
Compute 11
Compute 09
Compute 14
Compute 12
Compute 12
Compute 12
Compute 12
Compute 10
Compute 08
(21) Cabled Ports
(8) Node Fan-In Ports
(33 of 36) Ports Used
Notice the rich >3 ratio of ‘Fabric Ports’ to ‘Node Ports’ {(21+4)/8}
Closest like competitor has 1.25:1 (SUN 6048). Most have 1:1
SGI ICE 8400 has 2.4x to 3x the link to node bandwidth fabric interconnect
Company Confidential July 2011
46. Topologies
Hypercube topology
– Cables together switches in blade enclosures
– No external switches, less cables reduces costs
– Lower bi-section bandwidth per node, but also lower networking costs
– Fits well in some application spaces, but not all
– Suited for…
– larger node count MPI jobs
– low per node bandwidth requirement (i.e. less than 500MB/s/node)
– “nearest neighbor” type MPI communication patterns
– Constant Bi-Section (CBB) Fat Tree
– Common cluster topology
– High bi-section bandwidth per node
– Requires external switches
– As system size grows, cost of network grows rapidly
– Suited for…
– smaller node count MPI jobs
– high per node bandwidth requirement (i.e. GB/s/node or greater)
– “all-to-all” type MPI communication patters
Company Confidential July 2011
47. SGI Altix ICE- Industry Breakthrough
Compute Rack Level ‘Live’ Integration
NASA Ames Post - NAS TECHNICAL HIGHLIGHTS February 8, 2010
'Live' Integration of Pleiades Rack Saves 2 Million Hours (excerpt)
The new 512-core rack arrived in late December and installation was completed in early January.
Integration into the Pleiades system was accomplished by connecting the new rack's InfiniBand (IB) dual
port fabric via 44 fibre cables-while Pleiades was running a full production workload.
This live integration saved 2 million hours in productivity that have previously been lost each time a
planned system outage occurs. When outages on Pleiades are planned, users get a one-week notice and
system utilization plummets about three days before the actual shutdown. This drop in usage is partly due
to the fact that batch jobs are only started if they can finish by the start of the planned outage. About half
of Pleiades' computational hours are consumed by long-running jobs-most take five days to complete-
further adding to the usage slowdown.
http://www.nas.nasa.gov/News/TechHighlights/2010/2-8-10.html
•SGI’s superior hypercube based IB network topologies not only enables adding nodes and
switches but also now enables adding racks of nodes and switches without disturbing the
existing production load.
•Competitor network topology offerings such as fat tree and 3D torus are either
inherently limited or strictly incapable of supporting such a dynamic reconfiguration.
Company Confidential July 2011
48. Tightly Integrated Software Stack
Linux® Operating System
Performance
SGI® ProPack™ software for Linux OS
Optimization
System Management SGI® Tempo
Workload Manager Altair® PBS Professional™
MPI Intel MPI Runtime or SGI MPT
IB Fabric and Subnet SGI InfiniBand Fabric Subnet Management (based on
Management OFED and OpenSM)
Customizable, integrated solution stack
Cost-effective, standards-based & optimized for ease of use
Factory integrated and tested
Company Confidential July 2011
49. Customers That Are… "Our new SGI Altix ICE system was up and
running the afternoon it arrived, and we
Breaking Barriers with ICE began running benchmarks within 48 hours"
Matthew Bate, Professor of Theoretical
Astrophysics at the University of Exeter.
Company Confidential July 2011 49
50. Customers That Are… “Just as Columbia has helped NASA achieve
breakthroughs that were previously impossible, this new
Breaking Barriers with ICE supercomputer will enable NASA to continue tapping the
far limits of science and innovation.“
Dr. Rupak Biswas, acting chief of the NAS Division
Company Confidential July 2011 50
51. Customers That Are… “It quickly became clear to us that for the same
number of processors, the performance of the SGI
Breaking Barriers with ICE Altix ICE system was in a league of its own.”
Henrik Diamant, Head of CFD, Honda Racing F1 Team
Company Confidential July 2011 51
52. OS Noise Synchronization – SGI® Tempo™ Feature
• OS system noise: CPU cycles stolen from a user application by the OS to do
periodic or asynchronous work (monitoring, daemons, garbage collection, etc).
• Management interface will allow users to select what gets synchronized
• Performance boost on larger scales systems
Process on: Unsynchronized OS Noise → Wasted Cycles
System Wasted Wasted
Node 1 Overhead Cycles Cycles
Node 2 Wasted System Wasted
Compute Cycles Cycles Overhead Cycles
Node 3 Wasted Wasted System
Cycles Cycles Overhead
Barrier Complete
Process on:
System
Node 1 Overhead
System
Node 2 Overhead
System
Node 3 Overhead
Synchronized OS Noise → Faster Results
Slide 52
Company Confidential July 2011 Time
56. Staying in Front of the Data Gap
1982 1983 1987 1991 1992 1993 1996 1999 2001 2003 2005 2006 2007 2008 2009 2010 2011
SGI’s core focus for 25 years has been to push the leading edge of
efficiently creating & managing massive amounts of data
Company Confidential July 2011 56
57. Storage Platforms for Technical
Computing
InfiniteStorage JBOD Platforms InfiniteStorage Server Platforms InfiniteStorage RAID Platforms InfiniteStorage MAID Platforms Tape Libraries
Half-depth
Half- Half-depth
Half-
Half-depth Entry Level
Entry Level COPAN 400T/TX SpectraLogic Tape
SpectraLogic Tape
Half-depth COPAN 400T/TX
•Optimized for low cost •Optimized for low cost
•Optimized for low cost •Optimized for price
•Optimized for price •VTL •T50e
•T50e
•Optimized for low cost •VTL
capacity capacity
capacity performance
performance •Ground breaking density •T120
•T120
capacity •Ground breaking density
•2U and 3U enclosures •3U servers
•3U servers •Designed to limit power •T200
•T200
•2U and 3U enclosures •Designed to limit power
•2.5” or 3.5” disks
2.5” or 3.5” disks •3.5” disks
3.5” disks Enterprise Mid-Range
Mid- •T380
•2.5” 3.5” •3.5” Enterprise Mid-Range and cooling
and cooling •T380
•Integrates into a a wide •T680
•Integrates into wide •T680
Full-depth
Full- Full-depth
Full-
Full-depth •Performance, reliability
•Performance, reliability variety of environments •T950
•T950
Full-depth variety of environments
and features
and features •T-Finity
•T-Finity
•Optimized for
•Optimized for •Optimized for scalability
•Optimized for scalability COPAN 400M
COPAN 400M
performance •3U and 4U Servers
•3U and 4U Servers HPC
HPC •Strong feature set, like
•Strong feature set, like
performance
•2U enclosure •3.5” disks
3.5” disks
•3.5” •Disk to Disk Solutions integrated encryption,
integrated encryption,
•2U enclosure •Disk to Disk Solutions
•2.5” disks
2.5” disks •High performance and
•High performance and •DMF and popular media health monitoring
media health monitoring
•2.5” •DMF and popular
scalability
scalability applications and easy maintenance
and easy maintenance
applications
NAS – File Serving InfiniteStorage Shared Filesystems InfiniteStorage Software Suite
ArcFiniti CXFS DMF LiveArc
CXFS
•Integrated, network •Enables “online” access to “offline” data
online” offline” •Digital asset and knowledge
•High performance shared
•High performance shared
accessible Disk-based
Disk- •Dramatic cost reductions in large management platform
file system for Scale-Up
Scale-
file system for Scale-Up
archiver leveraging MAID environments •General (collaborative) application
SGI File Serving
Lustre
Lustre SGI InfiniteStorage Hardware Platforms
•Rules-based migration, enables fine
Rules-
tuning of storage resources
platform
•Components can be used to create a
diverse range of applications
•Price/performance file •High performance parallel
•High performance parallel
file system for Scale Out
file system for Scale Out
serving over IP or IB
Company Confidential July 2011 57
58. Storage Platforms for Technical
Computing
InfiniteStorage JBOD Platforms InfiniteStorage Server Platforms InfiniteStorage RAID Platforms InfiniteStorage MAID Platforms Tape Libraries
Half-depth
Half- Half-depth
Half-
Half-depth Entry Level
Entry Level COPAN 400T/TX SpectraLogic Tape
SpectraLogic Tape
Half-depth COPAN 400T/TX
•Optimized for low cost •Optimized for low cost
•Optimized for low cost •Optimized for price
•Optimized for price •VTL •T50e
•T50e
•Optimized for low cost •VTL
capacity capacity
capacity performance
performance •Ground breaking density •T120
•T120
capacity •Ground breaking density
•2U and 3U enclosures •3U servers
•3U servers •Designed to limit power •T200
•T200
•2U and 3U enclosures •Designed to limit power
•2.5” or 3.5” disks
2.5” or 3.5” disks •3.5” disks
3.5” disks Enterprise Mid-Range
Mid- •T380
•2.5” 3.5” •3.5” Enterprise Mid-Range and cooling
and cooling •T380
•Integrates into a a wide •T680
•Integrates into wide •T680
Full-depth
Full- Full-depth
Full-
Full-depth •Performance, reliability
•Performance, reliability variety of environments •T950
•T950
Full-depth
•Optimized for
SGI InfiniteStorage Software & File Systems
•Optimized for scalability
and features
and features
COPAN 400M
variety of environments
•T-Finity
•T-Finity
•Optimized for •Optimized for scalability COPAN 400M
performance •3U and 4U Servers
•3U and 4U Servers HPC
HPC •Strong feature set, like
•Strong feature set, like
performance
•2U enclosure •3.5” disks
3.5” disks
•3.5” •Disk to Disk Solutions integrated encryption,
integrated encryption,
•2U enclosure •Disk to Disk Solutions
•2.5” disks
2.5” disks •High performance and
•High performance and •DMF and popular media health monitoring
media health monitoring
•2.5” •DMF and popular
scalability
scalability applications and easy maintenance
and easy maintenance
applications
NAS – File Serving InfiniteStorage Shared Filesystems InfiniteStorage Software Suite
ArcFiniti CXFS
CXFS DMF LiveArc
•Integrated, network •Enables “online” access to “offline” data
online” offline” •Digital asset and knowledge
•High performance shared
•High performance shared
accessible Disk-based
Disk- •Dramatic cost reductions in large management platform
file system for Scale-Up
Scale-
file system for Scale-Up
archiver leveraging MAID environments •General (collaborative) application
•Rules-based migration, enables fine
Rules- platform
Lustre
Lustre
SGI File Serving tuning of storage resources •Components can be used to create a
diverse range of applications
•High performance parallel
•High performance parallel
•Price/performance file
file system for Scale Out
file system for Scale Out
serving over IP or IB
Company Confidential July 2011 58
59. SGI InfiniteStorage Software & File Systems
Aligned to Client Requirements
For clients requiring: For clients requiring: For clients requiring: For clients requiring:
• HSM • Digital Asset Management • High bandwidth Access to • Scalable FS to scale-out
Scale-up SSI servers clusters
• Transparent multi-tier data • Data Classification
migration • Real-time, streaming • Scalable to 100+ GB/s
• Tools to enable collaboration
• A methodology to “green” across disparate data workloads • Linux only
their storage environments • DMF back-end • HPC Environment
• Minimize administrative • Workflow management • IB, GigE or 10 GigE
overhead
• DMF back-end
DMF LiveArc CXFS Lustre
• Enables “online” access to • Digital asset and knowledge • Instant multi-OS, multi- • Medium to huge clusters
“offline” data management platform
platform, no copy, data with greater than 4GB/s
• Dramatic cost reductions in • General (collaborative) sharing requirements
large environments application platform
• Enables multiple file • Best solution for very
• Rules-based migration, • Components can be used to
enables fine tuning of storage create a diverse range of systems in a cluster large clusters
resources at file and volume applications • High availability with • Open source solution
levels • Extensible service-oriented automatic failure
• Maintains data in online state architecture (SOA) • Object-based file system,
detection and recovery
during migration requires object storage
• XODB, XML-based Object
• Partial-file migration for rapid Database server
access • Written in platform neutral Java
Company Confidential July 2011 59
60. SGI® InfiniteStorage Hardware
Platforms
JBOD InfiniteStorage RAID Platforms Persistent Data Stores
Entry RAID Mid-Range
Mid- Performance MAID
MAID Tape Libraries
Tape Libraries
Entry RAID Mid-Range Performance
• IS1000 Series
Half Depth JBOD • IS4600 • IS6120 • COPAN 400M • Spectra Logic
• IS5000 - Performance RAID Native MAID
- Mixed Use RAID - T50e
- Entry RAID - SSD, SAS, SATA drives
• IS2000 Series - FC or IB HBAs • COPAN 400T/TX - T120
- 6 Gb/s SAS intermix
Full Depth JBOD - Optional 4U VTL - T200
- SAS and/or SSD drives - 120 drives max
60-drive dense tray - T380
- 192 drives max
- 480 drives max • IS15000 - T680
• IS4100 - Extreme bandwidth, Integrated Archive
Integrated Archive - T950
• IS5500 - T-Finity
- Mid-Range RAID density & scalability
ArcFiniti
- FC and SATA drives - Next Gen RAID - 600 drives per rack
- 112 drives max - 6 Gb/s SAS - 1200 drives max • Network-accessible
Network-
- FC, IB, SAS Disk-based archive
Disk-
- SAS and/or SSD • IS16000
solution.
drives - 10GB/s Reads and
writes • Up to 1.44 PB in a
- 384 drives max
- 300k Random HDD single rack.
read IOPS • Leveraging MAID
InfiniteStorage Server Platforms
InfiniteStorage Servers
InfiniteStorage Servers
• ISS3009 • ISS3500
- Half-depth, 3U - 4U – 36 Drives
- 9 Drives - IB, GigE & 10GigE
• ISS3012 - SAS, SATA & SSD
- Half-depth, 3U - ISER, iSCSI, NFS, CIFs
- 12 Drives
• ISS3118
- Full-depth, 3U
- 18 Drives
Company Confidential July 2011 60
61. Storage Strategy, where SGI Plays
SGI Sweet Spot: Larger Deals & Integrated, Productive
Solutions
Block-Based
1000 Transaction
Transaction
Processing
Processing
Intranet/
Intranet/
Extranet
Extranet Data
System files Data
OLTP Warehousing
Banking
Warehousing
Retail … Electronic
Commerce
Databases
Data Warehousing
Data Mining
Internet
Internet
IOPS (thousands)
Scientific
Scientific
100
Cloud
Cloud Computing
Computing
Imaging
Imaging
10 Video
Video
File-Based
5 10 50 100 500
Throughput in MB/sec MB/sec
Company Confidential July 2011 61
62. SGI® InfiniteStorage Server 3500
An solutions platform with integrated compute and storage
– High-density, aggressively-priced storage server
– Single & Dual-socket Intel® Xeon® processors
– High-Availability configuration
• Redundant (1+1) 1400W Gold Level power supply with PMBus function
• 7 x 8cm (middle) hot-swap cooling fans
– 4U - 36 hot-swappable 3.5” drives – SAS, SATA & SSD
Company Confidential July 2011 62
63. Copan Technology
A Strategic Portfolio Enhancement
Significant savings in power and cooling
– All disks can be powered off when not in use
– Up to 50% of the disks can be powered on
Reduced floor space requirement
– Lower power requirements and reduced heat
generation allows high density packaging
Extends disk life
– DiskAerobics software
• Data integrity verification
• Proactive sparing
– Disk power down reduces wear and tear*
*Testing showed 3-6x reduction in failure rate (increased MTBF)
Company Confidential July 2011 63
64. Purpose-built Persistent Data
Platform
Advantages over transactional storage: Advantages over tape storage:
6x better drive reliability 20x faster data access
6x higher capacity in a single footprint Integrated with tape vaulting
32x better data protection than standard RAID RAID protection benefit
Unmatched product life of 7+ years Simplified media management
Assured drive health and data integrity Integrated with HSM data movers
Persistent data lives in every tier
Availability Reliability Footprint
Performance Data Protection Scalability Affordability
Tier I Tier II Tier III Tier IV Tier V
Enterprise Disk Modular Array Object-Based File Backup/Recovery Off-Site DR Vaulting
Mission Critical, Repurposed Disaster Recovery,
Low-Cost SATA Near-line Recovery
Transactional Storage Transactional Storage Off-Site Data
Company Confidential July 2011 64
65. Different Types of Data Demand Different
Storage Solutions
Transactional or Persistent Data Vaulted Data
Dynamic Data Data Protection and Archive Data
I/O intensive • Large files • Sequential • Offsite
Small files • Very large storage • Explosive growth • Copy of copy
Modest storage growth • Throughput • Sequential
Steady growth rates • Compliance
E-mail Document Database Backup Replication Maps Video Imaging
8KB 80KB 8MB 10MB 20MB 60MB 300MB 48GB
Transactional data Persistent data Vaulted data
Amount of data in the typical enterprise
Within 30 days the
Within 30 days the
majority of transactional
majority of transactional
data becomes
data becomes
persistent data
persistent data
Company Confidential July 2011 65 65
66. SGI COPAN 400 Platform Details
Disk-Based Core Platform
– Enterprise 1TB, 2TB or 3TB Drives
– 1 to 8 MAID Shelves
– Up to 2,688TB raw storage per cabinet with 3TB
drives
– FC 8Gbps MAID, FC 4Gbps VTL GigE 1Gbps
Performance
– Up to 6,400 MB/s, 23TB/hr. (native MAID) or
– Up to 3,200 MB/s, 12.5TB/hr.* (VTL)
Multiple Solutions
– Native MAID: ideal for HSM and D2D applications
- SGI Data Migration Facility - IBM TSM D2D
- Commvault Simpana - Quantum StorNext
– VTL: reliable, high performance target for backup
and archive applications
*speed to double in future software release
66
Company Confidential July 2011 66
67. Enabling Efficient Density
ArcFiniti’s Canister Technology
Exclusive mounting scheme to eliminate
“rotational vibration” within a storage shelf
Unique canister technology provides
even thermal distribution to reduce
cooling needs.
Industry-leading packing density
896 SATA drives in single footprint
2,688TB with 3TB drives
Company Confidential July 2011 67
68. Introducing ArcFiniti™
File-based Archiving Solution
Primary
– Virtualized storage tiers for scalability, lower cost. Cache
– TCO advantages with industry-leading power efficiency
and density featuring COPAN MAID technology Archive
Policy
– High-performance file-based access for easy integration
into existing infrastructures
Archive Tier
– Data protection software to ensure very long term data
integrity
– 1.4PB Usable archive in a single rack
--An easy-to-deploy, integrated archive solution--
Bringing together the best of SGI Storage Tech
Company Confidential July 2011
69. Objective:
Providing a true archive for all persistent file data
Document File System
management Media & Entertainment
E-mail Archive Archives
Software
Bio Sciences and Pharma
CCTV & Video
Surveillance
• Single platform supporting multiple
applications simultaneously
• Standard NFS File System Access
• CIFS SAMBA (V2)
• Hugely scalable, reliable and secure storage
Company Confidential July 2011 69
70. SAN without CXFS…
LAN
LAN
SAN
-- Each host owns dedicated volume in shared array
-- Files must be copied for sharing
Company Confidential July 2011 70
71. SAN with CXFS… True Storage
Consolidation
SAN LAN
-- Each host shares multiple volumes
-- Eliminates multiple copies of shared files
Company Confidential July 2011 71
72. Related Solutions:
DMF: Automated, Policy-Based Tier Virtualization
By replacing a Tape Library with an SGI COPAN System, you can significantly decrease access times
By replacing a Tape Library with an SGI COPAN System, you can significantly decrease access times
of even the oldest data in your DMF environment while maintaining cost optimization
of even the oldest data in your DMF environment while maintaining cost optimization
Company Confidential July 2011 72
73. SGI® InfiniteStorage 5000
Price-Performance Leader
Next-generation 6Gb/s SAS controllers
Three interface options
– SAS, FC/SAS, iSCSI*/SAS 12-Bay - 3.5-in Drive Tray
Up to 192 drives –
High capacity and nearline SAS, SSD** drives
– Three enclosure options
2 GB of cache per controller
– Mirrored, battery-backed, de-staged to flash 24-Bay - 2.5-in Drive Tray
Base and Hyper Performance options
Premium features
SnapCopy (8/64), VolumeCopy (8), Remote Volume
Mirroring (8 over FC ports)
*iSCSI available only in IS5000-SP
**SSDs and require a premium feature purchase
60-Bay - 2.5 & 3.5-in Drive Tray
Company Confidential July 2011 73
74. SGI® InfiniteStorage 5500
High performance, modular versatility
The next-generation RAID platform with Intel Jasper Forest
multi-core based SAS RAID Controllers
Two interface options – FC or Infiniband
– Eight 8Gb FC host ports standard (no HICs)
12-Bay - 3.5-in Drive
– Sixteen 8Gb FC ports with optional FC Host Interface Cards (HICs)
Enclosure – Four 40Gb Infiniband ports with optional IB HICs
Up to 384 drives per IS5500 array
– High Performance SAS, Value SAS, High Capacity SAS, FDE and SSD drives
– 2U24 or 2U12 or 4U60 drive enclosures – mix and match drive types
6 or 12 GB of cache per controller
24-Bay - 2.5-in Drive
Enclosure
– Mirrored, battery-backed and destaged to flash memory
Base and Turbo Performance options
Telco Ready - NEBs level 3 Certified & -48v DC Power Option*
Premium Features
– Snapshot Copy, Volume Copy, Remote Mirroring, Data Assure and Safe Store
Safe
60-Bay - 2.5 & 3.5-in
Drive Enclosure
Company Confidential July 2011
75. Converged SAN and NAS
SAN-attached clients
SAN-attached clients
Windows
•Leverage SGI’s XFS
•Leverage SGI’s XFS
Linux RHEL Linux SLES Server Windows® OS X 32 OS X 64 bit
•Seamless integration with
•Seamless integration with
near-line storage for unlimited
near-line storage for unlimited
data growth
data growth
•Data access via FC SAN, GigE,
•Data access via FC SAN, GigE,
10GbE, Infiniband via CXFS, NFS
10GbE, Infiniband via CXFS, NFS
SAN
SAN and CIFS
and CIFS
Primary
Storage
Secondary
Storage
(Optional)
Cluster
Cluster
NAS
NAS
MAID or
Tape
Library
Integrated Tiered Storage Network-Attached clients
Network-Attached clients
Integrated Tiered Storage
Company Confidential July 2011 75
76. Thank you
Martin Jalový, Systems Engineer SGI
mjalovy@sgi.com
Company Confidential July 2011