DevPlatform Party (by Yandex Infrastructure) 4 from 2023.10.17

It was interesting yesterday. mitap Devplatform by Yandex in Belgrade 4 speeches of varying degrees of density and about which I would like to tell

1. What is Capacity Planning Actually?
2. Parallel development and testing of microservices through feature branches
3. Where did the engineering romance go?
4. Boost infrastructure from scratch if you are a small startup

What is Capacity Planning Actually? (Report by Dmitry Nesterov of Yandex) In fact, I decided to watch Mitap because of this talk, since it tells about a topic that is relevant for large IT companies, or rather about planning and distribution of capacities. I was interested in the way Yandex in this topic, as we are on a similar route inside:) There were stages in Yandex. Appearance of capacity planning in 2016-2017 There are no clear processes and rules, no automation, but there is excel.) Developing capacity planning in 2018-2019 The emergence of a systematic approach (applications, capacity planner role, communication with customers) Development of planning capacity in 2020-2021 Technology problems have been solved, automation has appeared, but it is inconvenient. Maturity of capacity planning in 2022-2023 - introduced a product approach (We went from JTBD users.)Formalized and simplified processes, there is a development strategy Then Dmitry showed and told how these tools work and it was interesting.

Parallel development and testing of microservices through feature branches (Dmitry Kostyukov from Alfa Bank) If Mitap had started with this report, I wouldn't have watched it. The story is not about the quality of the report, but about the approach to technology and platform development. In fact, Dmitry talked about how to test features in the conditional process of gitflow with a bunch of microservices lined up in a chain. This problem he solved through his own way...

NOIn developing a technology platform, it is important to make the right things easy to do and the curved things difficult to do. And in software development right now, the target code process is TBD. (trunk based development)which does not require long-lived branches. To achieve it, you need to write a lot of automated tests, be able to work with feature toggles and so on. This allows you to quickly develop and quickly see and resolve conflicts in the logic of features, as it is visible inside the code. (It's not like in gitflow where you see it during merge.). In such a scheme, all this machinery that was mentioned in the report is not necessary, and even harmful, since it provokes to make long-lived branches. As a result, the tactical task of the guys from the platform Alfa Bank decided, but the strategic ... alas:)

Where did the engineering romance go? (report from Yaroslav Astafyev) Philosophical history, which was interesting to listen to and which should motivate development. In principle, it was interesting to listen, but this is more of a software report.

Boost infrastructure from scratch if you are a small startup (report from Alexey Shagraev from pora.ai) Lesha's story about using AWS Fargate (Serverless compute for containers) To organize infrastructure for a startup:) And a simple CI/CD to get there. The bottom line is that startups are cool to use as a Service solutions that allow you to immerse yourself in creating business logic instead of building an infrastructure that provides basic abstractions and knows how to self healing. It seemed to me that this is a fairly simple story that Lesha told, because he once worked at Yandex:)

#SoftwareDevelopment #PlatformEngineering #Engineering #SRE #Devops #Architecture