If you are buying Designing Data-Intensive Applications today, buy the second edition. If you already own the first, it is not obsolete, and for interview preparation specifically it still covers what interviews actually test.
That is the short answer. The longer answer matters if you are deciding whether to pay again, hunting for a secondhand copy, or wondering whether the new material changes how you should prepare. This article covers what changed, what stayed, and how either edition fits into interview preparation.
What the second edition is
The second edition of Designing Data-Intensive Applications was published in February 2026, nine years after the original. Martin Kleppmann wrote it with a co-author this time: Chris Riccomini, an engineer and author who built data infrastructure at LinkedIn.
It is a full revision rather than a patch. The whole book was reworked for clarity and for a decade of technology change, and it runs about 60 pages longer than the first edition.
What actually changed
The additions cluster in three areas.
Data systems for AI. The first edition predates the current wave of AI systems. The second adds coverage of vector indexes, which are the data structures behind semantic search, plus DataFrames and the batch systems used to prepare training data.
Cloud-native architecture. The first edition was written when running your own database was the default. The second covers the architectures of managed cloud data systems, which is how most readers now meet these ideas in practice.
A sharper framing around trade-offs. The second edition opens with a chapter on trade-offs in data systems architecture and adds a chapter on defining nonfunctional requirements. The chapter on consistency and consensus, the hardest material in the book, was almost completely rewritten for clarity.
| First edition | Second edition | |
|---|---|---|
| Authors | Martin Kleppmann | Kleppmann and Chris Riccomini |
| Length | About 590 pages | About 60 pages longer |
| AI-era data systems | Not covered | Vector indexes, DataFrames, training pipelines |
| Cloud-native systems | Pre-cloud framing | Covered directly |
| Consensus chapter | Famously dense | Almost completely rewritten |
| Core concepts | Replication, partitioning, transactions, batch, stream | The same, revised for clarity |
What stayed the same
The core of the book is the same body of knowledge. Replication, partitioning, transactions, consistency models, batch processing, and stream processing did not change as concepts in the intervening decade, and those chapters are revised rather than replaced.
That is why the first edition is not obsolete. A secondhand first edition still teaches the fundamentals correctly. What it lacks is the AI-era material, the cloud framing, and the clearer rewrite of the hardest chapters.
Does the new edition change interview preparation?
Mostly no, with one interesting exception.
System design interviews test fundamentals: how replication fails, when to partition, what consistency costs. Both editions cover that material, so owning the first edition puts you at no real disadvantage in an interview.
The exception is the new framing. The second edition now opens with trade-offs and nonfunctional requirements, which is precisely the shape of an interview conversation. An interviewer does not ask you to recite how consensus works. They ask what your design gives up, and what requirements justify the cost. The first edition contains that judgment implicitly. The second edition puts it in the front of the book.
The standing caution still applies to both editions: DDIA is a textbook, not interview prep, and reading it cover to cover in the weeks before an interview is a common mistake. Which chapters help and which to skip is covered in DDIA for system design interviews.
Which parts to read before an interview
Read by topic, not by chapter number, because the chapter numbering differs between editions:
- Replication and partitioning. The highest-value material in the book for interviews. How leaders fail, how followers lag, how data splits across machines.
- Transactions and consistency. Enough to say what you lose when you relax consistency, and why.
- The trade-offs and nonfunctional requirements chapters in the second edition. They map directly onto the first ten minutes of a design interview.
- Skip for now: the encoding internals, the deep storage engine material, and the batch and stream chapters, unless the role is data engineering.
Pair the reading with practice on real questions. The free case study lessons on Design Gurus apply the same concepts to systems like YouTube and Dropbox, which is the format the interview actually uses.
Is there a free PDF?
Searches for a second edition PDF are common, so here is the honest answer. There is no legitimate free PDF of either edition. Copies floating around on GitHub or file-sharing sites are pirated, often incomplete, and sometimes altered.
The legitimate ways to read it without buying the print book: an O'Reilly online subscription includes it, some employers and universities provide O'Reilly access already, and public libraries often carry the print edition. A secondhand first edition is also cheap now that the second has replaced it on shelves.
Frequently asked questions
Is the first edition of DDIA obsolete? No. The fundamentals it teaches are unchanged and still correct. It is missing the AI-era data systems material, the cloud-native framing, and the rewritten consensus chapter.
Is the second edition worth it if I own the first? For interview preparation alone, no. For your work, yes if you touch AI data systems or want the clearer consensus rewrite. Otherwise the first edition you own still serves.
When was the second edition published? February 2026, by O'Reilly. Martin Kleppmann was joined by co-author Chris Riccomini.
Is DDIA good interview preparation? It is the best book for understanding distributed systems and a poor tool for short-term interview prep. It is organized by topic, not by interview question. Use selected chapters over a long horizon, and use interview-shaped material when the interview is near.
How long does DDIA take to read? Cover to cover, most engineers report a few months of steady reading. The interview-relevant subset, replication through consistency, is a few weeks.
Should I read DDIA or take a course? Different jobs. DDIA builds deep understanding over months. A course like Grokking the System Design Interview organizes the same fundamentals around the interview format in about 20 hours. Doing the course first also makes DDIA a faster read.
Related reading
- DDIA for System Design Interviews: What to Read and What to Skip
- Grokking the System Design Interview Alternatives: 7 Options Compared
- Grokking the System Design Interview vs Alex Xu's Books
- Top 30 System Design Interview Questions and Answers
- A Beginner's Guide to System Design Interviews
The second edition is the better book, and the first edition is still a correct one. For interviews, either works as the deep reference behind your preparation, and neither replaces practicing on interview-shaped questions. If you want the interview-shaped half assembled already, that is what Grokking the System Design Interview is for.