From 0 to 25M Users: Lessons in Scaling Backend Services Over Six Years

6 years ago – Launched the system in 3 days and 3 nights

When the application finally ran smoothly, the sense of time seemed to freeze, marking the birth of a new service.

5 years ago – First capacity expansion

Within a year the system grew to hundreds of thousands of users, exposing bottlenecks that prompted a move to cluster load balancing.

4 years ago – Over 5 million cumulative users

The rapid user growth brought a lingering sense of unease, as increasing concurrency pressure threatened system stability.

3 years ago – "Brazil World Cup" event overload

The high‑traffic event exposed severe concurrency and architectural issues; seven days of continuous maintenance and tuning could not hide the system’s inefficiency, and temporary scaling proved only a short‑term fix.

2 years ago – Full service‑oriented refactor

Reflecting on the "World Cup" incident revealed that the root cause lay both in the system and in personal limitations. Inspired by emergent behavior and distributed thinking, the monolithic application was decomposed into independent services. After months of core framework development, the system handled 25 million cumulative users and 8 million monthly active users with zero failures.

1 year ago – Sustainable service framework

Embracing the principle "keep it simple" led to continuous efforts to shorten the path from problem to solution, focusing on essential simplicity. Even under 20 million monthly active users and high concurrency, the system remained stable while the author kept pursuing simplicity, efficiency, flexibility, and deeper understanding.

Today – Ongoing curiosity and micro‑ecosystem vision

Curiosity and self‑reflection drive the next step: building a "micro‑ecosystem" where micro‑services, like individuals in an ecological system, can intelligently sustain themselves. The goal is to support 650 k daily active users and over 100 k concurrent requests while continuing to explore and innovate.