Without Surprise.

Volume II · Free online edition

Further Reading

Timothy O’Brien 6 min read Free online

These are the works I would keep close while doing the investigations in this book; the chapter notes point to the specific claims, tools, and equations used. The SRE book, both editions of Forecasting: Principles and Practice, Quantitative System Performance, and the framework and methodology pages are free online; the other books are sold in print and through subscription libraries. Where my emphasis differs from theirs, the question is usually what happens when a useful method meets several businesses, technologies, and authorities at once.

J.R. Storment and Mike Fuller, Cloud FinOps: Collaborative, Real-Time Cloud Value Decision Making, second edition (O’Reilly, 2023). The second edition added chapters on forecasting and on partnering with engineering to its treatment of cost allocation, rate discounts from commitments, and unit economics. Read it alongside the FinOps Foundation’s evolving framework, which the Foundation itself describes as flexible. Financial accountability, collaboration, and business value are common ground. My emphasis is on choosing the level at which engagement is useful: direct work with consequential consumers, central action for consumers without control, and several operating arrangements within one organization.1

Arun Kejariwal and John Allspaw, The Art of Capacity Planning: Scaling Web Resources in the Cloud, second edition (O’Reilly, 2017). The authors present a practical approach rather than theoretical models. Measure current capacity, because without measurement “you would only be guessing”; turn the measurements into forecasts through a process you can adjust; then let those forecasts shape deployment and autoscaling. Carry that discipline into a setting where requirements can change faster than supply, a service can become critical unexpectedly, and the organization needs several planning methods at once.2

Neil J. Gunther, Guerrilla Capacity Planning: A Tactical Approach to Planning for Highly Scalable Applications and Services (Springer, 2007), and his later Universal Scalability Law materials. The law’s two parameters, contention and coherency delay, are fitted to throughput measured at several load levels; the coherency term explains why adding resources can reduce throughput. Quantitative models like this deserve serious attention. So do their tested ranges, measurement error, and the mechanisms that invalidate an extrapolation. Volume I, Chapter 5 applies that distinction to changed data and workload behavior.3

Edward D. Lazowska, John Zahorjan, G. Scott Graham, and Kenneth C. Sevcik, Quantitative System Performance: Computer System Analysis Using Queueing Network Models (Prentice-Hall, 1984), and Mor Harchol-Balter, Performance Modeling and Design of Computer Systems: Queueing Theory in Action (Cambridge University Press, 2013). The first supplies the operational laws, among them the utilization law and Little’s law, which test whether a set of measurements is consistent before anyone fits a model to it. The second develops the queueing results behind response times that climb steeply as utilization approaches saturation, and the difference between open and closed systems. Volume I, Chapter 5 uses both on load tests and core counts, and Volume I, Chapters 9 and 10 apply Little’s law to agents and GPU serving.4

Betsy Beyer, Chris Jones, Jennifer Petoff, and Niall Richard Murphy, editors, Site Reliability Engineering: How Google Runs Production Systems (O’Reilly, 2016). The introduction counts demand forecasting and capacity planning among an SRE team’s responsibilities. The chapter on software engineering in SRE describes Auxon, which generates resource allocation plans from an encoded statement of each service’s dependencies and requirements; the authors say many teams at Google moved to this intent-based planning, which is not evidence that every team there, or at any other enterprise, works this way. Volume I, Chapter 3 develops why planning authority can extend well beyond the people operating a live service.5

Microsoft’s Azure Well-Architected Framework. Read the Cost Optimization and Reliability guidance together, particularly their discussions of tradeoffs. They connect spending choices with recovery, redundancy, scaling, and operating effort; the cost guidance explicitly warns against choosing solely for a lower bill.6 I would use those questions with the workload specialists before presenting an optimization, then bring the resulting requirements into the portfolio discussion. That gives the architecture review and the capacity plan a common set of conditions to defend.

Rob J. Hyndman and George Athanasopoulos, Forecasting: Principles and Practice, third edition (OTexts, 2021). The freely available text is the mathematical foundation I would keep beside the forecasting work: simple benchmarks such as seasonal naive, evaluation on a rolling forecasting origin, judgmental adjustments, prediction intervals, and forecast reconciliation. Its examples use R; Forecasting: Principles and Practice, the Pythonic Way (2026) carries the same material into Python with Nixtla’s open-source libraries. Chapters 14–16 adapt these ideas to decisions whose supply lead times, customer promises, and evidence dates matter.7

Brendan Gregg’s performance methodologies and the technologies’ own documentation. Gregg’s methodology pages and his Systems Performance: Enterprise and the Cloud, second edition (Addison-Wesley, 2020), provide a disciplined starting point: characterize the workload, then check every resource for utilization, saturation, and errors with the USE method. Continue with engine-specific and serving-specific evidence: Cassandra maintenance and placement, Kafka delivery and replay, PostgreSQL query and recovery behavior, and AI serving and evaluation. Volume I, Chapters 5–12 cite the primary sources and separate documented capabilities from measurements; none of the database, load, or AI benchmarks they describe was run for this book.8

Three books from outside the field are worth the detour. None mentions a cloud bill, and each trains the habit this book depends on: looking past the artifact to the system that produces it.

Donella H. Meadows, Thinking in Systems: A Primer, edited by Diana Wright (Chelsea Green, 2008). A short, clear account of stocks, flows, feedback loops, and delays, and of why the most obvious lever is usually the weakest. Its ranked list of places to intervene in a system, from parameters up through rules, information flows, and goals, is a good companion to the Part IV argument that information flows, rules, and goals move a system further than spending targets do.

Rick Rubin, The Creative Act: A Way of Being (Penguin Press, 2023). A record producer’s book about attention and reduction, with nothing in it about engineering. Read it for its discipline of noticing, and for the test the Introduction borrows: how many pieces can you remove before the work stops being the work?

Douglas R. Hofstadter, Gödel, Escher, Bach: An Eternal Golden Braid (Basic Books, 1979). A long, playful book about how self-reference and meaning arise from formal systems, told through mathematics, art, and music. Its cover, a pair of blocks whose shadows are the letters G, E, and B, gave the Introduction its picture of cost, capacity, and capability as three views of one system.

Read these alongside the system you are investigating and the people who understand it. A good reference can help you ask a question you would otherwise have missed; the answer still has to come from the workload and the business in front of you.

Notes and sources

  1. O’Reilly, Cloud FinOps, second edition, and the publisher’s description of its new chapters; FinOps Foundation, FinOps Framework. Checked October 3, 2026; no later edition found.↩︎

  2. O’Reilly, The Art of Capacity Planning, second edition, September 2017: publisher’s description (“rather than theoretical models”) and the chapter openings of “Measurement: Units of Capacity” and “Predicting Trends”. Checked October 3, 2026.↩︎

  3. Springer, Guerrilla Capacity Planning, copyright 2007; Neil J. Gunther, “How to Quantify Scalability”, which describes diminishing returns from contention and negative returns from incoherency. Checked October 3, 2026. The chapter’s limits on extrapolation are the author’s application of these foundations.↩︎

  4. Edward D. Lazowska, John Zahorjan, G. Scott Graham, and Kenneth C. Sevcik, Quantitative System Performance, Prentice-Hall, 1984; the authors’ page says the publisher returned the copyright and offers the full text to view or download. Mor Harchol-Balter, Performance Modeling and Design of Computer Systems, Cambridge University Press, February 2013. Checked October 3, 2026. Chapter 5’s notes cite the chapters used.↩︎

  5. Google, Site Reliability Engineering, 2016: Chapter 1, “Introduction” and Chapter 18, “Software Engineering in SRE”, which describes intent-based capacity planning as an approach “many teams have moved to.” Checked October 3, 2026. The enterprise governance argument is this book’s analysis.↩︎

  6. Microsoft Learn, “Cost Optimization tradeoffs” and “Reliability tradeoffs”, Azure Well-Architected Framework, accessed September 30, 2026, and rechecked October 3, 2026. The cost page says that “choices that focus only on minimizing spending can undermine your workload’s business goals and reputation.” Vendor architecture guidance describing consequences across workload requirements, not a measured result for this book’s portfolio.↩︎

  7. Rob J. Hyndman and George Athanasopoulos, Forecasting: Principles and Practice, third edition (OTexts, 2021); Hyndman, Athanasopoulos, Azul Garza, Cristian Challu, Max Mergenthaler, and Kin G. Olivares, Forecasting: Principles and Practice, the Pythonic Way (OTexts, 2026), which describes itself as based on the third edition. Checked October 3, 2026. The chapter notes cite the specific sections used.↩︎

  8. Brendan Gregg, Performance Analysis Methodology, The USE Method, and the official page for Systems Performance, second edition. Checked October 3, 2026.↩︎