5 Ways Companies Are Using SUSE HPC in AI, ML and analytics
Download as PPTX, PDF1 like321 views
The document discusses how companies utilize SUSE high-performance computing (HPC) for artificial intelligence (AI), machine learning (ML), and analytics across various industries, highlighting its significance in sectors like healthcare, finance, and manufacturing. It emphasizes SUSE's role in providing a robust HPC platform that enhances the development and management of AI/ML solutions, despite not having its own AI/ML offerings. The document outlines specific use cases and benefits of integrating HPC across different domains, while noting the growing relevance of HPC beyond traditional scientific research.
5 Ways Companies Are Using SUSE HPC in AI, ML and analytics
- 1. 5 Ways Companies Are Using SUSE High Performance Computing in AI, Machine Learning and Analytics Jeff Reser Product & Solution Marketing SUSE / jeff.reser@suse.com
- 2. 22 Prologue – Setting the Stage
- 3. 3 HPC market factoids HPC ROI is very high - $463 (on average) revenue per dollar; $44 average profit (or cost savings) per dollar invested in HPC1 Worldwide HPC revenue expected to reach over $19.5 billion by 20221 Big data combined with HPC creating new solutions, adding many new users/buyers to the HPC space (AI/ML/DL and HPDA are hot new areas) SUSE runs on nearly half (24) of the top 50 supercomputers (5 RH, 7 CentOS)2 SUSE dominates top 100, CentOS gains share in “smaller” supercomputers2 Commercial OS Share in Top 500 (represents 111 supercomputers in the list): SUSE 59%, RH 22%, bullx 15%, Ubuntu 4%2 1 Hyperion Research, November 2018 2 Top500 Supercomputer Report, November 2018
- 4. 4 Where does SUSE HPC fit? SUSE does not have an AI/ML solution, but delivers an HPC platform for parallel computing that makes those solutions better (easier to develop, faster and more manageable) Businesses around the world today are recognizing that a Linux-based HPC infrastructure is vital to supporting the analytics applications of tomorrow High Performance AI, Machine Learning, Internet of Things and High Performance Data Analytics are powered by High Performance Computing and fueled by big data HPC is not just for scientific research any longer, and is being adopted across banking, healthcare, utilities and manufacturing
- 5. 5 SLE HPC product factoids HPC Module with supported HPC packages – beyond an OS Supports Aarch64 (Arm) and x86-64 architectures Many IHV/ISV/CSP partnerships Actively investing in HPC Competitively priced
- 6. 6 SUSE Package Hub Kubernetes, micro-services and containers OpenACC and accelerators Software-defined storage for HPC Arm 64 support gaining popularity HPC in the Cloud HPC module HPC-related projects
- 7. 7 SUSE HPC Reference Architecture Clustering Tools Storage Interface
- 8. 8 Artificial Intelligence Machine Learning Neural Networks Deep Learning Convolutional Neural Networks Transfer Learning The Spectrum of AI Solutions Deep Learning Examples are disease identification and energy demand optimization. Machine Learning Examples are cyber security, autonomous vehicles and F1 racing. Artificial Intelligence Examples are Google Maps and game play. Neural Networks Examples are facial and voice recognition. Convolutional Neural Networks Examples are image/video recognition and medical image analysis. Transfer Learning For example, knowledge gained while learning to recognize cars could apply when trying to recognize trucks.
- 9. 9 Rise of High Performance AI • HPAI is containerized, many small files in a hyperscale environment with data transfer within a single node (PCI bus) • Traditional HPC distributes compute over many nodes, smaller number of large files with data transfer between multiple nodes
- 10. 10 Banking & Finance • Risk modeling in financial portfolios • Fraud detection in real time as transactions are processed between disparate systems • Cybercrime • High-freq trading applications • Pricing / regulatory compliance Security & Safety • Smart cities and power grids • Surveillance • Image analysis • Facial recognition and detection Retail • Affinity marketing • Smart inventory & supply chain management • Cyber fraud • Image tagging • Speech recognition • Sentiment analysis Automotive & Transportation • Automotive modeling and aerodynamic design • Engine performance and timing, fuel consumption • Safety systems • Self driving, AI driverless operations • Logistics and mapping Healthcare • AI for precision and personalized medicine • On-demand diagnoses, diagnostic assistance and treatment plans • Drug discovery and research • Cancer research and cancer cell detection • Remote surgeries Energy / Oil & Gas • Air flow • Heating and cooling efficiencies • Wind turbines • Ocean exploration • Seismic imaging • Reservoir characterization • Sub-surface fault detection Manufacturing • Computational Fluid Dynamics • Heat dissipation systems • AI advanced robotics • Automated systems • High performance designs AI/HPC use cases are popping up across industries
- 14. 14 Key ISVs
- 15. 1515 Scene I – Consumer Goods
- 16. 16 Consumer Goods – Appliance Design Goals • Reduce time in the development of innovative ovens, washing machines and refrigerators • Simulate/test more products/variations in a virtual design space before committing to a physical prototype Challenges • Find the design “sweet spot” within complex customer, safety, regulatory and cost requirements • Reduce customer returns and associated revenue loss incurred from products and packaging damaged in shipment • Decrease costs and time-to-market delays associated with physical prototyping
- 17. 17 Consumer Goods – HPC + AI/ML • Saves time and money because prototypes are used for validation, not development • Improves customer satisfaction with higher quality products that include more customer design features • Saves money because packaging and product design innovations substantially reduce shipping damage (fewer returns) • AI/ML applied across the entire product and service cycles – developing leaner, more optimized operations and personalized customer experiences • Companies are benefiting by achieving higher sales, reducing manufacturing costs and improving the retail brand.
- 18. 18 And there’s more … • Affinity marketing – customer buying behaviors, analysis of buying patterns, precise ad targeting and what products to suggest to repeat customers • Smart inventory and supply chain management – product trends, determining what to stock, determining what to stock and how much to buy • Cyber fraud – credit card fraud, maintaining data privacy and combating cybercrime
- 19. 19 Consumer Goods – ISV Solutions Solutions • Create virtual design spaces, where engineers can model advanced design solutions (including multiple requirements) and run these more complex simulations faster • Simulate full products and corresponding packaging as an integrated “system” that withstands shipping rigors and arrives undamaged Vendors • ANSYS structural mechanics software • Engineers may do multiple simulations in in a day instead of spending weeks building a prototype – the company can now evaluate many more design alternatives and get better product to market faster • Engineers can put multiple physical systems together in a virtual test to make better product tradeoffs – products can be optimized for energy efficiency or noise reduction or extra consumer features.
- 20. 20 Many simulation types from electronics to fluid flow to structures and systems plus multi- physics More than 50 ANSYS workflows certified on SLE for engineering simulation across the entire product lifecycle, from digital exploration to operations and maintenance • Racing aerodynamics testing and optimizations • Integration testing for complex, distributed, smart systems in modern jet aircraft • Development of fully-electric super sports car
- 21. 2121 Scene II – Energy & Utilities
- 22. 22 Energy & Utilities – Sustainable Energy Goals • Diversify and increase pace of innovation • Improve organizational and infrastructure efficiencies • Increase responsiveness by aligning product supply • Achieve safe and compliant operations Challenges • Increasing health/safety/environmental regulations and public scrutiny • Manage complexities and risk of aging infrastructure and new technologies • Changing priorities and demands of customers and society
- 23. 23 Energy & Utilities – HPC + AI/ML • Simulations and modeling help reduce cost and risk of future operations • AI/ML/DL revolutionizing the way we produce, transmit and consume energy • AI/ML/DL helps limit environmental impacts and optimizes supply/demand • Energy smart grids powered by AI/ML, continuously collecting overwhelming amounts of data from millions of smart sensors nationwide to make timely decisions on how to best allocate energy resources
- 24. 24 And there’s more … Hyperloop • Tube systems for passengers and vehicles rushing at speeds up to 700 miles/hour • Magnetic accelerators and compressed air bearings remove frictional forces • Design anticipates operational maintenance and variables such as earthquakes, power outages and passenger fluctuations • Design must be safe, reliable, affordable and self-powered. • HPC simulations enable iterative design changes and show results
- 25. 25 And even more … Tidal energy • Tidal turbines use an undulating membrane that “swims” in the moving current • Tidal waters move the membrane and specialized electronics convert the motion into usable energy • Tidal energy is more efficient than solar or wind energy, produces no pollution or waste, no visual impact, no noise, no harm to wildlife and does not impede navigation • HPC simulations tests virtual models to ensure performance, practicality and cost-effectiveness before manufacturing physical prototype that meets sustainability targets
- 26. 26 Energy & Utilities – ISV Solutions Solutions • Create consolidated View of Energy Plants -- the impact of potential changes can be validated to meet the original requirements • Expedite renewable energy development -- such as biomass, water, wind, solar, fuel cells, tidal and wave, and hydropower plays an ever-increasing role in the global share of electricity generation Vendors • Dassault Systemes provides virtual representations of physical plants and systems, enabling validation and verification of plant design, construction safety procedures, operations and even personnel training. • ANSYS enables engineering and research organizations to better develop new renewable energy equipment. From testing for the electrochemical performance of a fuel cell stack to optimizing the design of biomass reactors and photovoltaic collectors, ANSYS solutions help speed the development process and bring renewable technologies rapidly to market.
- 27. 27 Simulation to explore real-world behavior, delivered on the 3DEXPERIENCE platform • Collaborate, model and optimize operations • Consumer experience and marketing effectiveness • Data management for smarter decision-making • Planet modeling and simulation
- 28. 2828 Scene III – Automotive
- 29. 29 Automotive – Design Goals • Design, model, test and refine control strategies in an integrated design environment • Use processes that reduce design times, enabling engineers to create and test their ideas quickly and with few hardware prototypes • Provide a seamless environment to take designers’ ideas from concept, through verification to real product code Challenges • Speed up design, increase quality and reduce R&D costs by finding an alternative to traditional design methods • Bridge the gaps in traditional automotive electronics development
- 30. 30 Automotive – HPC + AI/ML • HPC provides the best platform for running design and simulations along with the tools used in the design processes • AI/ML assists with advanced safety features for connected vehicles • AI cloud platforms ensure data is available when needed • AI speeds up the process of filing claims when accidents occur • Car manufacturing optimized in innovative new ways • Driving monitoring enhanced by AI, detecting and monitoring driver behavior, recognition and infotainment control
- 31. 31 Automotive – ISV Solutions Solutions • Coordinate within the electronic control units (ECUs) – the under-dash controllers that run the software that runs the vehicles. • Focus on performance for fuel injection and transmission controls to meet emissions standards while delivering improved performance, Vendors • MathWorks tools have found a real home in ECU development • Synopsys provides tools and IP to help manufacturers create the special chips needed to address special needs in automotive systems • Specifications for powertrain controllers, for example, now begin in the intuitive and self-documenting environment of Simulink and Stateflow, both powered by the industrial-strength computational, analytical and visualization capabilities of MATLAB.
- 32. 32 MathWorks is the leading developer of mathematical computing software for engineers and scientists Its universal ECU and embedded coder facilitates rapid prototyping for many car types Synopsis delivers the world’s most advanced tools for silicon chip design, verification, IP integration and application security testing, certified and supported on SUSE • Automotive design and test • Delivery of next generation IoT devices • Cutting edge silicon chip design
- 33. 3333 Scene IV – Manufacturing
- 34. 34 Manufacturing – Materials Goals • Build very strong, yet lightweight materials that can withstand extreme temperatures • Exceed safety and compliance regulations for aerospace, deep space, automotive, oceanic and more Challenges • Developing materials requires deep understanding of their underlying properties, and that requires time-consuming computational processes on a massive scale • Need an intuitive, easy to use and effective way of managing clusters and not take valuable time away from research
- 35. 35 Manufacturing – HPC + AI/ML • HPC modeling helps discover materials faster • AI/ML predicts which compounds can create materials before setting foot in a lab • Algorithms can predict which compounds combine to form interesting new materials • AI can mine databases for “recipes” to make the materials • Modeling helps quickly map out exactly what makes a material so much stronger or lighter
- 36. 36 Manufacturing – ISV Solutions • Bright Cluster Manager manages, provisions and monitors the HPC clusters, allowing researchers to focus on their work instead of managing IT infrastructure • Bright Computing provides full support for HPC via GPUs, from drivers that render GPUs programmable via CUDA, to the CUDA-enabled tools, to CUDA-optimized libraries critical to materials science research • Univa Grid Engine is a leading workload management solution that provides scheduling capabilities to AI/ML and GPU applications. • System admins have complete visibility and control over every aspect of their clusters and gain productivity advantages through the tight integration of the two products. • Clusters are automatically cloud-ready, enabling system administrators to extend their clusters into the cloud, and manage these nodes, with just a few easy steps.
- 37. 37 Bright Cluster Manager delivers outstanding cluster management with quick deployments for data science readily done Bright Cluster Manager lets users monitor and build clusters of any size that are easy to provision, operate, monitor, manage and scale, all visible on one pane of glass. Univa Grid Engine is the leading distributed resource management system that optimizes resources by transparently selecting the resources that are best suited for each segment of work. Grid Engine software manages workload placement automatically, maximizes shared resources, supports extreme scale, provides enterprise-grade dependability and accelerates deployment of any container, application or service in any technology environment, on-premise or in the cloud. Bright and Univa work together closely to help HPC users set up and manage their systems including their cloud images while smoothly provisioning and managing their workloads across the HPC space.
- 38. 3838 Scene V – Healthcare
- 39. 39 Healthcare – Pharmaceuticals & Drug Research Goals • Faster drug discovery, design and validation • More predictable drug treatment results among patients • Consolidated view of servers, clusters and instances in the public cloud Challenges • Analyze and integrate different types of biomedical and healthcare data • Integrate public cloud into the existing IT infrastructure with minimal administrative overhead, and no disruption to user workflow • Find new and naturally occurring compounds from food sources rather than synthetic compounds
- 40. 40 Healthcare – HPC + AI/ML • ML predicts treatment results, matching drug interactions with patients • ML/DL in drug design, discovery and origins virtually combines atoms to generate new molecular structures • Combine AI and DNA analyses to discover ingredients in different food sources that have therapeutic qualities, including the management of chronic metabolic diseases. • AI initiatives in precision medicine and health data applications
- 41. 41 Healthcare – ISV Solutions Solutions • Create treatments for gene-specific diseases based on better understanding of the genome • Track and analyze patient outcomes across whole healthcare systems to identify best practices and share them. • Ensure consistent and to-spec manufacturing of medical devices Vendors • Altair PBS Pro, they configured each of the four geographically distinct locations as peers from the workload-management perspective. PBS Pro maintains a real-time inventory of node availability on a per-location basis. Jobs submitted to PBS Pro are separated into six types based upon resource requirements that detail job size, processing and memory requirements, urgency, software image and other needs. • Bright Cluster Manager, seamless integration with the public cloud can be achieved. The cloud can be easily used as an additional resource on the system with virtually no administrative overhead. • SAS analytics helps hospital groups gather and analyze very large amounts of patient data to identify and promulgate best practices, improving healthcare on a wide scope
- 42. 42 Altair makes HPC faster, smarter & easy to manage with PBS Works™ Altair provides services for applications that streamline the workflow management of compute- intensive tasks including solvers, optimization, modeling, visualization and analytics • Simulation technology to synthesize and optimize designs, processes and decisions • Accurate, reliable vehicle component modeling • Management and coordination of HPC development • Innovative drone propulsion design
- 43. 43 SAS is a market leader for multi-variate analyses, business intelligence, data management and advanced predictive analytics SAS just announced a $1B investment in AI capabilities, R&D and education • Business intelligence • Advanced analytics • Fraud and security intelligence
- 44. 4444 Epilogue
- 45. 45 We examined 5 industry scenarios … Energy & Utilities Sustainable energy Healthcare Pharmaceuticals & drug research Manufacturing Materials Automotive Design Consumer Goods & Retail Appliance design
- 46. 46 … and validated some conclusions • SUSE does not have an AI/ML or GPU solution, but delivers an HPC platform that makes those solutions better (easier to develop, faster and more manageable) • Businesses around the world today are recognizing that a Linux-based HPC infrastructure is vital to supporting the analytics applications of tomorrow • HPC is not just for scientific research any longer, and is being adopted across banking, healthcare, utilities and manufacturing • Key ISVs for SUSE HPC environments:
- 47. 47 Questions? Just for fun, if you figured out my puzzle just send the solution to jeff.reser@suse.com
- 48. SUSE LINUX HPC POWERS AI ML DL GPU WORKLOADS
- 49. 4949 Unpublished Work of SUSE LLC. All Rights Reserved. This work is an unpublished work and contains confidential, proprietary and trade secret information of SUSE LLC. Access to this work is restricted to SUSE employees who have a need to know to perform tasks within the scope of their assignments. No part of this work may be practiced, performed, copied, distributed, revised, modified, translated, abridged, condensed, expanded, collected, or adapted without the prior written consent of SUSE. Any use or exploitation of this work without authorization could subject the perpetrator to criminal and civil liability. General Disclaimer This document is not to be construed as a promise by any participating company to develop, deliver, or market a product. It is not a commitment to deliver any material, code, or functionality, and should not be relied upon in making purchasing decisions. SUSE makes no representations or warranties with respect to the contents of this document, and specifically disclaims any express or implied warranties of merchantability or fitness for any particular purpose. The development, release, and timing of features or functionality described for SUSE products remains at the sole discretion of SUSE. Further, SUSE reserves the right to revise this document and to make changes to its content, at any time, without obligation to notify any person or entity of such revisions or changes. All SUSE marks referenced in this presentation are trademarks or registered trademarks of Novell, Inc. in the United States and other countries. All third- party trademarks are the property of their respective owners.