WIXLIB

Building an artificial intelligence (AI)information architecture (IA) with IBM Storage

Table of contents01 IntroductionArtificial intelligence (AI) is a journey that begins withdata. Thus, AI cannot exist without an informationarchitecture (IA). Being able to gain business valueand insights from data is not always easy. Legacyinfrastructure is inadequate for AI workloads and datasilos make it difficult to get a holistic view of all yourinformation, limiting the value of AI.02 There is no AI without y, organizations are moving towards hybridcloud to respond to evolving business needs. As datais increasingly distributed, it becomes a struggleto provide adequate protection and management.Infrastructure that was not built for AI and hybridcloud is not flexible enough to respond to modernworkloads and demands without adding complexity.03 Modern IA for AI andhybrid cloudThe best AI is built on a foundation of data that iscollected, organized and analyzed carefully, theninfused into the business. This foundation shouldalso be open, flexible and allow access to data ofevery type, regardless of where it lives.IBM Spectrum ScaleIBM Cloud Object StorageIBM Spectrum DiscoverIBM Elastic Storage SystemEvery successful AI project goes through amultistep process that starts with having theright data and progresses to using AI broadly.04 Case studies: Creatinga competitive advantageContinental AutomotiveUniversity of Birmingham05 Conclusion76% of IT decision-makerssurveyed said AI will be a key part of their digitaltransformation strategy over the next one to two years.121 . IDC White Paper, sponsored by IBM, Accelerating AI Modernization with Data Infrastructure, doc # US47460721, February 2021.2

01 IntroductionAdopting AI is not without its challenges. Thegeneral-purpose storage infrastructure thatorganizations are accustomed to using needs tobe replaced or supplemented with storage systemsthat are geared towards AI specific tasks.1 Moving from experiments to scaling AI forbusiness value. Modernizing for an AI-focuseddigital transformation requires expertise innew standards of developing, implementingand maintaining AI solutions at scale. Legacy infrastructure/complexity. Organizationscan no longer use traditional, general-purposecomputing or legacy storage infrastructure. Thisoutdated infrastructure increases complexityand is not flexible enough to respond to AIworkload demands. Instead, they must employa scalable compute with an equally scalable,high-performing, integrated, flexible and securestorage infrastructure. Data silos. Storage is typically implementedwith specific storage solutions that create silosof data. These silos are not integrated together,nor are they integrated with a comprehensiveset of infrastructure solutions, resulting in a lackof global data access.The AI LadderIt is no surprise many organizations are not surehow to proceed and do not have a clear understandingof how best to leverage AI to their advantage. Thisis why IBM has put together a prescriptive approachto accelerating the journey to AI called the AI Ladder .The AI Ladder is a framework that accelerates yourability to collect and organize data, gain deeperinsights by leveraging AI-driven data analysis andinfuse your enterprise with these insights.Organizations face a few core challengeswhen adopting AI, including scaling AIfor business value, the use of legacyinfrastructure and elimination of data silos.31 . IDC White Paper, sponsored by IBM, Accelerating AI Modernization with Data Infrastructure, doc # US47460721, February 2021.

02 There is no AI without IAAIIAThere isno AIwithout IAAs companies begin to modernize, theyseek to provide an architecture that willpropel them into the future. The journeyto AI is about moving data from ingestto insights with an IA that can easily beintegrated throughout the organization.It is important that each part of the AILadder provides an integration to the entirejourney. Starting a project on one part ofthe journey is fine, but it is critical to ensurethe project considers an overall IA for AIto optimize resources and modernize yourinfrastructure for expanding AI workloads.CollectAnalyzeData is the fuel that powers AI, but it can become trapped orstored in a way that makes it difficult or cost-prohibitive tomaintain or expand. You will need to unleash that data so itcan expand from edge to core to public cloud within a simpleand cost-efficient infrastructure. IBM Storage for data and AImakes data simple and accessible for hybrid cloud with AIstorage solutions that fit your existing business model.Analysis is critical to the AI journey and must provide high performancefor fast analysis and seamless connection to both the data lake andthe storage catalog. Organizations must plan for issues beyond thedeployment of AI; you need to build AI infrastructures that give youconfidence in your data and that enable you to access it wherever itresides. IBM Storage for data and AI provides high-performanceaccess to data and an integrated AI infrastructure for analysis.OrganizeAI can only be as good as the data it relies on. Businessesmust fully understand what data they have so they canleverage it for AI and other organizational needs, includingcompliance, data optimization, data cataloging and datagovernance. IBM Storage for data and AI provides insights intodata from multiple sources by automatically and continuouslyindexing objects and files when changes are made and storingthis information in the built-in storage catalog.InfuseBusiness challenges can become an opportunity to explore,understand, predict and bring an AI infrastructure to everyorganization. IBM Storage for data and AI is empowering youto use data and AI storage to leverage that infrastructure inmore ways that bring value to your organization.ModernizeA solid IA is the foundation for AI and hybrid cloud. Modernizingyour infrastructure means building a foundation that not only takesadvantage of cloud-native technologies, but also drives AI throughoutthe organization. IBM Storage for data and AI delivers the flexibilityneeded to respond to AI workloads, and integrates with Kubernetesand the Red Hat OpenShift platform, making it easier to deploycloud-native applications.4

03 Modernize IA for AI and hybrid cloudModernize IA for AIand hybrid cloudAI initiatives are easier and more likelyto succeed if they are built on a solidfoundation. IBM Storage for data andAI provides that foundation with acollection of offerings that modernizeyour IA and address the top businesschallenges associated with deployingAI workloads.IBM Spectrum ScaleIBM Cloud Object StorageIBM Spectrum DiscoverIBM Spectrum Scale is a highly scalable,data-efficient, high-performance storagesolution with enterprise security and aglobal parallel file system for both file andobject storage data. IBM Spectrum Scaleenables the unification of data across ahybrid cloud into a single scale-out storagesolution for the entire data center fromedge to core to public cloud. IBM SpectrumScale is available both as a software-onlysolution or as an integrated appliance.IBM Cloud Object Storage is a highlyscalable cloud storage solution forunstructured data that provides onpremises and cloud-based dedicatedservices. IBM Cloud Object Storage usesan innovative approach for cost-effectivelystoring large volumes of unstructureddata. It delivers the capabilities requiredto provide continuous access to dataassets while improving research outcomes,decision making, business responsivenessand regulatory or legal demands.IBM Spectrum Discover is a multisource datacatalog that automatically and continuouslyindexes objects and files whenever changesare made using the metadata. It can also beused to create custom tags and policy-basedworkflows to orchestrate content inspectionand activate data in AI, machine learning(ML) and analytics workflows. IBM SpectrumDiscover helps enable faster AI analysis,compliance classification, image and videoindexing, personal data identification, AI datapipeline integration, real-time data discoveryand new insights to optimize data and findbad or duplicate data.Learn about IBM Spectrum ScaleLearn about IBM Cloud Object StorageLearn more about IBM Spectrum Discover5

03 Modernize IA for AI and hybrid cloudIBM Elastic Storage SystemIBM Elastic Storage System (ESS) is a modern implementation ofsoftware-defined storage, making it easier to deploy fast, highlyscalable storage for AI and the hybrid cloud.Learn about IBM Elastic StorageIBM Elastic Storage System 5000IBM Elastic Storage System 3200IBM Elastic Storage System 5000(ESS 5000) is designed for data lakeswith increased performance, density andscalability. With ESS 5000, you canconsolidate massive data volumes,increase simplicity and accelerate speed.IBM Elastic Storage System 3200(ESS 3200) is designed to meet and beatthe challenge of managing data for analytics.Learn about IBM ElasticStorage System 3200Learn about IBM ElasticStorage System 50006

04 Case studies: Creating a competitive advantageContinental AutomotiveResultsAccelerating insight intovehicle safety at Continental150 yearsContinental has pushed theboundaries of automotiveinnovation for 150 years.For many people, driving is simply a series ofautomatic decisions. Training AI to make thosesame decisions even a 10th of a second fasterrequires petabytes of data. To develop autonomousdriving solutions that potentially make driving safer,Continental used IBM Elastic Storage System,IBM Spectrum Scale and NVIDIA DGX systems to:70%Continental improved AItraining time 70% usingIBM Spectrum Scale andNVIDIA DGX systems. Modernize its application development withoutgiving up on infrastructure requirements likeperformance, scalability or simplicity. Ensure that its infrastructure will support the growthrequired, whether in the cloud or on premises. Optimize for deep learning with multi-node training,enabling it to increase model accuracy for higherlevels of safety without impacting time to production.Read the case study“The collaboration betweenContinental, IBM Storage andNVIDIA is bringing a promiseto life in terms of safety.”Robert ThielHead of AI, AdvancedDriver Assistance,Continental Automotive AG14xContinental has the abilityto run at least 14x moredeep learning experimentsper month at the same time.7

04 Case studies: Creating a competitive advantageUniversity of BirminghamResultsDriving innovative researchforward by taking control of dataSupports compliance with dataprotection regulations at lowcost and without disruption.Today’s research simulations generate more datathan ever before. To meet this ever-increasingdemand, the University of Birmingham deployedIBM Spectrum Scale and IBM Spectrum Protect to: Provide a single data management planeacross multiple storage systems. Enable price-performance decisions whenmatching workloads to platforms, withoutcausing complexity to spiral out of control. Allow researchers to deploy applications whereit makes sense with immediate data availability.Read the case study“We support research in a wide range of areas,including applying and developing techniquesto use AI and deep learning. For example, we’recollaborating with the University of Nottinghamon the Centre of Membrane Proteins and Receptorsproject. By analyzing the super high-resolutionimages produced by the latest generations ofmicroscopes, the project will shed light on howcardiovascular disease, respiratory disordersand cancer can be better prevented and treated.”Simon ThompsonResearch ComputingInfrastructure Architect,University of BirminghamUp to 2Up to 2 FTEs estimatedsaving due to enhancedoperational efficiency.5,0005,000 researcherssupported byinfrastructure that helpsthem find solutions tokey issues faster.8