AI and Analytics at Scale

Book description

Getting large-scale data-driven applications for AI and analytics into production doesn't have to be challenging. Technical managers, senior technologists, and implementers today often overlook fundamental aspects of design and data infrastructure—aspects that can make the difference between failed approaches and reliable, successful production systems.

In this exclusive report, you'll learn which practices work—and which don't—at large and innovative companies that have successfully integrated AI and analytics into their workflows. Over the past two years, authors Ted Dunning and Ellen Friedman have worked with a wide range of businesses to deliver in-production systems at a large scale. You'll learn practices that have been particularly beneficial, including many that have been disregarded.

  • Understand why AI is at its best when coupled with analytics
  • Build successful production systems—running AI and analytics on the same infrastructure—at scale with less effort, pressure, and cost
  • Apply aspects of a scale-efficient system, including a comprehensive data strategy, containerization, and scalability without scaling IT
  • Focus on the increasingly popular topics of AI and edge computing
  • Explore an example data infrastructure: HPE Ezmeral Data Fabric

Table of contents

  1. Preface
  2. 1. Should AI and Analytics at Scale Be Difficult?
    1. Data Challenges at Scale
      1. Effective Data Storage
      2. Data, Kubernetes, and Containerized Applications
      3. In-Production Data from the Start
      4. Flexibility of Data Access
    2. What Makes Production Break?
      1. Security at Scale
    3. What Does Scale Really Mean?
      1. Scale in Terms of Data Size
      2. Scale in Terms of the Number of Files or Other Objects
      3. Scale Can Mean Many Applications or Many Teams
      4. Scale in Terms of Geo-distributed Locations
    4. Scalability Is as Important as Scale
      1. Proliferation as a Symptom
    5. Scale Up Without Scaling IT
    6. Look for Faulty Assumptions
    7. The Shape of the Solution
  3. 2. Scale-Efficient AI and Analytics
    1. Comprehensive Data Strategy
      1. What Is a Comprehensive Data Strategy?
      2. Counter Example: A Nonscale-Efficient Approach
      3. Universal Data Access
      4. Real-World Example: Data Warehouse Off-Load
      5. Real-World Example: Good Data Infrastructure Simplifies Steps
      6. Use Positive Incentives to Avoid Data Silos
      7. What Is in a Data Guarantee?
      8. Real-World example: Proving Data Guarantees
      9. Leaning Forward, but Looking Back
    2. Containerized Computation with Kubernetes
      1. Real-World Example: Containers Enable Smaller Footprint and Less IT
      2. Dealing with Data in a Kubernetes World
      3. Keep Legacy Applications from Weighing You Down
    3. Separation of Concerns
      1. Platform-Level Policies and Actions
      2. Real-World Example: Provisioning a New Data Center
      3. Self-Service Data and Application Management
      4. Real-World Example: Data Motion at Platform Level
    4. Plan for Scalability, Not Just Scale
      1. Growth Without Having to Re-architect Your System
      2. Stampede on Successful Data Infrastructure
      3. Real-World Example: Rapid Growth in User Demand
      4. Growth Without a Requirement to Scale IT
      5. Real-World Example: Responding to Seasonal Spikes
      6. Flexibility to Deal with Temporary Scale
    5. Multiuse and Multitenancy: Winning Strategies at Scale
      1. Who’s on the Other Side of the Wall?
      2. Collocate Work Based on Data, Not on Organizational Structure
      3. Real-World Example: Lunchroom Collaboration Yields Millions
      4. Real-World Example: Reuse Data Without Proliferation
      5. Real-World Example: Reuse Disaster Recovery Facilities
      6. Stepping Stones
  4. 3. AI and Analytics Together
    1. Why AI and Analytics Together?
      1. Second-Project Advantage
      2. Handling Logistics Efficiently
      3. A Human Advantage
      4. Real-World Example: Identifying Business Value
    2. Challenges and Solutions
      1. Flexibility Through Open Data Access
      2. Real-World Example: Traditional and Modern Tools Together
      3. The Coattail Tactic
      4. Real-World Example: Real-Time Response Through AI
      5. Real-World Example: Analytics in Production Plus AI in Development
      6. Containers and Kubernetes for Isolation and Customization
      7. Real-World Example: Containerization Improved AI Pipelines
      8. Data Versioning: Platform-Level Capabilities
      9. DataOps and IT: Bigger Impact, Less Effort
      10. Edge and AI Often Co-occur
  5. 4. AI and Analytics in Edge Systems
    1. Edge Means Many Things
      1. Geo-distribution
      2. Real-World Example: Telemetry Backhaul
      3. High-Volume Ingest
      4. Real-World Example: Autonomous Car Development
      5. Security and Ownership
      6. Edge Management
  6. 5. Example Data Infrastructure: HPE Ezmeral Data Fabric
    1. What Is HPE Ezmeral Data Fabric?
      1. Universal, Multiple-API Data Access
      2. Scalability, Reliability, and Performance
      3. Platform-Level Data Management
      4. Data Movement Without Interference
    2. Conclusion
  7. 6. Where to Go from Here
  8. Additional Resources

Product information

  • Title: AI and Analytics at Scale
  • Author(s): Ted Dunning, Ellen Friedman
  • Release date: January 2021
  • Publisher(s): O'Reilly Media, Inc.
  • ISBN: 9781492094371