Architecture Weekly #164
Architecture Weekly Issue #164. Articles, books, and playlists on architecture and related topics. Split by sections, highlighted with complexity: 🤟 means hardcore, 👷♂️ is technically applicable right away, 🍼 - is an introduction to the topic or an overview. Now in telegram and Substack as well.
System Design Course Cohort #5 is open!
Typical system design courses teach technical skills but often overlook the connection to business problems. This course fills that gap, emphasizing the importance of recognizing and addressing business priorities with technical approaches. Learn to go beyond load balancing options and performance tactics by focusing on solving real business challenges. Now - accompanied by the AI tools! Course was already completed by 50+ engineers with great feedback!
SIGN UP HERE: ONLY 2 slots left!
Highlights
Decomposing Transactional Systems 🤟
Top pick of this week is no doubt the post by Alex Miller destructuring how different databases like Google Spanner, FoundationDB, Tapir and others take 4 essential steps of executing, ordering, validating and persisting transactions. Phil Eaton contributed, Marc Brooker wrote a follow up with explaining Aurora SQL. Long story short: absolute must read!
#distributedsystems
On the Biology of a Large Language Model
Ever wondered what’s happening inside an AI’s “brain”? This paper plays mad scientist with Claude 3.5 Haiku (launched Oct 2024) and maps its internal circuits the way biologists trace neurons. The authors use attribution graphs—think MRI scans for transformers—to expose which nodes light up for tasks like syntax or world facts. The result feels half detective story, half lab report, and it gives fresh clues on why big models behave the way they do. Grab it if you like peeking under the hood instead of treating LLMs as magic boxes.
#ai #llm
How Much Should I Be Spending On Observability?
A bomb of a post from Charity Majors - again. We do lack that kind of posts where the technical aspect is considered from business perspective, and Charity does great job bringing it to the observability cost. 50% of your bill going to logs, observability cost growth and many more other insights in this great piece.
#observability
Follow-Up
Why I’m No Longer Talking to Architects About Microservices 👷♂️
Here’s a rant with a purpose: Ian Miell is done with endless “microservices” debates that never touch real business pain. He argues most architecture chats devolve into buzzword bingo, missing basics like cycle time, reliability, and cost. The piece nudges us to ditch holy‑war theology and focus on outcomes—break things up only if it genuinely speeds teams up or cuts failure blast‑radius. Expect blunt stories from conference rooms where eyes glazed over faster than your CI pipeline. A quick, spicy read to reset your next architecture meeting.
#microservices
Avoiding the Soft Delete Anti‑Pattern 👷♂️
Soft‑deletes sound cozy—just mark a row “dead” and never really kill it—but this essay calls them out as the database version of stuffing junk in a closet. It walks through the chaos they unleash on foreign keys, uniqueness checks, and query sanity, then parades four cleaner fixes: real deletes plus point‑in‑time recovery, explicit “archived” states, temporal tables, and off‑loading history to a warehouse. The tone is friendly but ruthless: stop lying to your RDBMS, it says, and design for the lifecycle you actually need. A fun read if you’ve ever slapped on a deleted_at column and hoped no one would notice.
#db
Isolation Levels & MVCC: A Comparative Field Guide 🤟
Franck Pachot turns the usual dreary ACID lecture into a world tour of Oracle, MySQL, SQL Server, Postgres, and YugabyteDB. He shows why “Repeatable Read” in one engine can act like “Snapshot” in another, and how MVCC rewrites the SQL‑92 rulebook. Expect live demos, not ivory‑tower theory, plus a cheat‑sheet for choosing the right isolation level when latency or write‑skew keeps you up at night. Perfect for anyone who’s nodded along in meetings while secretly googling “phantom reads.”
#db
Every Caching Strategy Explained in 5 Minutes 👷♂️
Caching is an essential tactic for software performance. This bite‑sized guide goes through Cache‑Aside, Read‑Through, Write‑Through, Write‑Behind, and Write‑Around caching strategies with tiny flowcharts and when‑to‑use tips. Think of it as speed‑dating with Redis patterns—enough to impress the interviewer without swallowing an O’Reilly book. Bonus: a quick quiz at the end lets you flex those fresh neurons.
#performance
SQLite File Format Viewer 🤟
Ever opened a .sqlite file and wished you had X‑ray vision? This fully client‑side web app lets you drag‑and‑drop a database and then walks every page, header, and B‑tree like a guided museum tour. Perfect for storage‑engine nerds, forensic dabblers, or anyone debugging weird WAL ghosts at 3 AM. Hacker News folks are already gushing about how it demystifies the “first‑page header squish” and other quirks. Runs completely in the browser, however I don't recommend upload something bigger than 100 MB.
#db
Crafting AWS Diagrams with Python + Cursor 👷♂️
Sick of pixel‑pushing in draw.io? This tutorial shows how to crank out multi‑account AWS diagrams using the Diagrams Python library while letting Cursor’s AI autocomplete the boring bits. It starts from a blank innovate_arch.py, adds clusters for Prod, Staging, etc., slaps color themes, and even throws in ArgoCD and CloudFront bling—then spits out a PNG on demand. Version‑controlled, reproducible, and guaranteed to keep your infra docs from rotting faster than bananas in July. Good excuse to tell your boss “it’s code, so it’s technical debt‑free.”
Big thanks to Nikita, Constantin, Anatoly, Oleksandr, Dima, Pavel B, Pavel, Robert, Roman, Iyri, Andrey, Lidia, Vladimir, August, Roman, Egor, Roman, Evgeniy, Nadia, Daria, Dzmitry, Mikhail, Nikita, Dmytro, Denis and Mikhail for supporting the newsletter on Patreon!