MLOps Literature

Operations work has a shelf of books behind it, and reading them in order turns a pile of tools into a way of thinking. The list starts with the twelve-factor app, then data-intensive systems, SRE, and distributed systems, moves to Jez Humble's delivery books, and ends with code craft: clean code, domain-driven design, a Python course, and a talk archive.

Twelve-factor

The first idea is how a single service should be built so that it can be operated at all. 12 Factors. The Twelve-Factor App is a methodology for building scalable and portable SaaS applications. It emphasizes best practices like maintaining a single codebase, isolating dependencies, storing configuration in the environment, treating services as replaceable resources, and enabling fast, stateless, and resilient deployments across consistent environments.

Data-intensive applications

Once one service is well built, the hard part moves to the data it reads and writes. Designing Data-Intensive Applications: reliable, maintainable is the book, and the Medium post is a Tech Book Talk review of it, part of a monthly series reviewing a technical book finished in the last month.

SRE

Data systems that people depend on have to stay up, which is what site reliability engineering is about. Google SRE is Google's page for its SRE books, a guide to SRE principles, best practices, and the role of a reliability engineer. The two books behind it are Site Reliability Engineering, the principles, and the Site Reliability Workbook, the practical companion.

Distributed systems and microservices

Reliability gets harder when the system is split across many machines and services. Designing Distributed Systems: patterns and paradigms is the book of distributed-system patterns, and Building Microservices: Designing Fine-Grained Systems is the book on cutting a system into fine-grained services.

Jez Humble

With the architecture in place, the next question is how changes get delivered through it, and Jez Humble's books are the delivery shelf. Jez Humble is his Amazon author page with the full bibliography. From it, Accelerate: software and high-performing organizations is about software and high-performing organizations, the DevOps Handbook is the DevOps book, and Continuous Delivery: deployment is the deployment book. Lean Enterprise completes the list, without a link.

Clean code

Delivery pipelines only help if the code moving through them is readable. Clean Code (Java). The first chapters are good. For Python, Clean Code in Python starts from the point that there is no strict definition or formal measure of clean code: checkers, linters, and static analyzers are necessary, but not sufficient. The Medium post is a book talk on that same Python book, and the Git repo is zedr's 🛁 Clean Code concepts adapted for Python.

Domain-driven design

Clean functions still need a model of the business they serve. Domain-Driven Design is the reference, with its definitions and summaries.

The Python course

For the language underneath most of this work, a single video course covers it end to end. The Complete Python Course is that video course.

pyVideo

After the course, talks keep the learning going. pyVideo is PyVideo.org, the archive of Python talks.