Software Engineering Social Connection: Yubo’s lean approach to scaling an 80M-user infrastructure

How did Yubo scale its infrastructure for 80 million users with a lean engineering team of just five?

#1about 2 minutes

Scaling Yubo's infrastructure for 80 million users

A small team of five engineers supports 80 million users by adopting a lean infrastructure philosophy and an MVP-first approach.

#2about 1 minute

The initial state of technical debt and complexity

The infrastructure began as a single, poorly optimized 2000-node GKE cluster with significant technical debt and a copy-paste culture.

#3about 2 minutes

Merging data engineering and DevOps for scalability

Internalizing the data engineering team within the infrastructure team enabled proactive database design and reduced production scalability issues.

#4about 4 minutes

Enabling developer autonomy with GitOps and CRDs

Moving from copy-pasted manifests to a GitOps model with ArgoCD, Helm templates, and custom Kubernetes operators empowered developers with self-service capabilities.

#5about 1 minute

Standardizing CI pipelines using Argo Workflows

CI pipelines were standardized using Argo Workflows, allowing developers to configure and launch their own pipelines by editing a simple YAML file.

#6about 2 minutes

Managing data consistency with change data capture

The "one database per feature" model is supported by a change data capture (CDC) pattern using Kafka to replicate data and keep services in sync.

#7about 2 minutes

Applying templates to ML and new application deployments

The same templating and CRD system used for applications is applied to simplify machine learning deployments and rapidly replicate the entire infrastructure for new products.

#8about 5 minutes

Building a cost-effective hybrid observability platform

A hybrid observability strategy combines SaaS tools like Datadog with open source solutions to control costs using dual shipping and metric compression patterns.