AI and Analytics at Scale

Chapter 2. Scale-Efficient AI and Analytics

We have said that you can build successful production systems—including running AI and analytics on the same infrastructure—at scale with less effort, less pressure on IT teams, and less cost than many people do. In this chapter, we show you how that is done. We will look at the style of scale-efficient AI and analytics systems, identifying fundamental aspects that make them successful and that you can put to work in your own organization. The most important of these are:

• Comprehensive data strategy

• Containerization of applications, orchestrated by Kubernetes

• Separation of concerns (supported by platform-level logistics)

• Scalability, not just scale

• Multiuse and multitenancy design

Solutions based on these fundamental aspects of scale-efficient AI and analytics systems are not just aspirational; we’ve observed (and sometimes helped build) them across a wide variety of businesses and a variety of sectors among our customers in the past several years. So we also describe in this chapter the patterns we’ve seen in real-world examples. In doing this, we’ll also examine how organizations take advantage of modern technologies to free themselves of limitations that would otherwise lead to the faulty assumptions discussed in Chapter 1.

Throughout the chapter, watch for the style of design and architecture that has been successful as well as the capabilities in data infrastructure and computational orchestration that enable these ...