Architecture Weekly Issue #183. Articles, books, and playlists on architecture and related topics. Split by sections, highlighted with complexity: 🀟 means hardcore, πŸ‘·β€β™‚οΈ is technically applicable right away, 🍼 - is an introduction to the topic or an overview. Now in telegram and Substack as well.

Business Oriented System Design Course Cohort #7 starts Wednesday at 30% off!

I have got an entire course to help you design software solutions and eventually pass interviews. 10 hours of content packed lectures, engaging practice and the final work you will be proud to showcase as well as Credly(by Pearson)-based digital certificate proving your experience. More than 100 engineers already passed the course with amazing feedback and advanced their careers. New cohort starts on 1st of October. 

APPLY NOW and grab limited offer of 30% OFF with the LASTCHANCE30 promocode!

Highlights

The Cost of Logging in 2022 πŸ‘·β€β™‚οΈ

You added winston library to your Node.js app; how much of a hit your througput took? Do you think 1%? More like 50%! Yes, logging is expensive. Although Matteo offers their own Pino library for logging, this post shows the real price you pay in terms of performance for observability.

The Cost of Logging in 2022 | Nearform
Nearform is an independent team of engineers, designers and strategists. We build digital capability and software solutions for ambitious enterprises seeking sustained business impact. We love what we do.

#observability #performance

Breaking free from rising observability costs πŸ‘·β€β™‚οΈ

Finally somebody broke down the explanation of the "You pay 3 times for the same observability data" notion which was discussed multiple times. See the evolution of observability tooling, 3 pillars of signals and how columnar dbs and OpenTelemetry unlocked the the solution to observability cost.

Breaking free from rising observability costs with open cost-efficient architectures
Discover how open, cost-efficient architectures can break the cycle of runaway observability costs and vendor lock-in

#observability #cost

Your AI Product Needs Evals πŸ‘·β€β™‚οΈ

Recently I was explaining a playbook to a set of managers how to run AI-driven projects(let me know if you're interested too!); feedback is a crucial part of it! Hamel Husain explains why ai results evaluation is so important and how to implement it.

Your AI Product Needs Evals –
How to construct domain-specific LLM evaluation systems.

#ai #qa

Follow-Up

Autoscaling Kubernetes Pods based on HTTP Traffic πŸ‘·β€β™‚οΈ

One of the performance tactic is horizontal scaling. However, it should rely on certain signals, and traffic is one of them. Grab a practical guide how to configure Horizontal Auto-Scaler in Kubernetes,

#kubernetes

PostgreSQL FOR UPDATE vs FOR NO KEY Update πŸ‘·β€β™‚οΈ

SELECT FOR UPDATE is the common mechanism to serialize update; but in some cases it can be too strict. Updating the post in a social network should not prevent you from writing comments to this post. For such a case there is SELECT FOR NO KEY UPDATE. Find the deep dive in this Vlad Mihalcea's post.

PostgreSQL FOR UPDATE vs FOR NO KEY UPDATE - Vlad Mihalcea
Learn the difference between the PostgreSQL FOR UPDATE and FOR NO KEY UPDATE when locking the parent record and inserting a child row.

#db #performance

The CTO's Todo list 🍼

One of the challenge of every technical manager is the work visibility; CTO is the ultimate case. Etienne de Bruin wrote an awesome piece on how you're seen by CEO, investors, clients and your own technical team and advise what to change in your behavior to change the narrative.

The CTO’s Todo List
The Four Visibility Multipliers That Create Legends

#cto

Data fabrics and data lakes πŸ‘·β€β™‚οΈ

In order to create a compliance-related report like DAC7 at Supplied we had to build a data architecture which allows to import data from different sources, combine them in a particular measure and then use for report generation itself. Apparently we were right in our architecture approach(no surprise though given the team πŸ˜„ ). Fivetran basically calls what we create a data fabric, where universal data storage is a corner stone.

Data fabrics and data lakes: One platform for universal data storage
How flexibility and interoperability can help you overcome vendor lock-in.

#dataarchitecture

Harvest, Yield and Scalable Tolerant Systems 🀟

The best practical application of CAP theorem that everyone sees literally everyday is the sacrifice of correctness and fulness of data in favor of availability. To really understand why it is so, read the paper on the Harvest, Yield and their measure in partial tolerant systems.

#distributedsystems #paper

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!