Scaling a Collaboration Service like Nextcloud to 20 Million users
summary
Nextcloud scales from a Raspberry Pi for a few users to a global deployment serving 20 million accounts by combining Kubernetes, Ceph, MariaDB clustering with new microservices and high‑performance backends. Production patterns include a clustered app tier behind load balancers, a global site selector that routes users to regional clusters, and federation to preserve shares across sites. Concrete results: Magenta Cloud runs 3.6M users on an OTC Kubernetes setup; Claro Drive reaches 20M with multi‑site clusters; face/object detection averages ~0.3s/image so one server would take ~10 years, 100 servers finish in weeks.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 3:10What is Nextcloud
- 11:43AI and machine learning features
- 16:59Scaling Nextcloud basics
- 20:12High performance backends
- 23:43Cluster setup
- 25:30Magenta Cloud case study
- 27:20Global scale architecture
- 33:20Summary and compliance
- 34:25Q&A session
related talks
12mScaling a Monolith, Is A Four-Letter Word
51mScalability Q&A
38mKubernetes Scalability: Federation & Cluster API
54mMigrating to Cloud Native with Microservices
19mKeynote: Operating 600+ Mesos Servers on 7 Datacenters @Criteo
35mGiving Private Clouds Public Capabilities
30mFrom 100 to 100M Users: How to Scale Your Cloud Architectures 🚀
32mScaling Up for Scalable Apps: The CIF-N26 Experience