Famous App Architectures Explained
Ever wondered what happens when you tap a reel, press Play or click Buy Now? These guides explain how the world's famous apps are built, in very simple English. Every fact about a real company is based on publicly shared information. Each guide also shows a clearly labelled illustrative design for building a similar app on AWS, with VPC, subnets, route tables, load balancer, EKS, S3, databases and monitoring.
How to read each guide
- Start with the example: what happens when you use the app.
- Look at the public architecture diagram and press Play to follow a request.
- Read which cloud provider, databases and tools the company publicly uses.
- Study the "If we built it on AWS" section. This is a learning design, not the company's real setup.
- Take the 4-question quiz at the end.
Quick comparison
| App | Cloud / hosting (public info) | Architecture style |
|---|---|---|
| Meta data centres (AWS until 2014) | Django monolith + Cassandra, PostgreSQL | |
| Netflix | AWS (4 regions) + own Open Connect CDN | Java microservices, GraphQL |
| Amazon.com | AWS | Service-oriented (100+ services per page) |
| Prime Video | AWS | Services + AWS Elemental media pipeline |
| Airbnb | AWS | Rails monolith to services on Kubernetes |
| Meta data centres | Hack/HHVM + thousands of microservices, TAO over MySQL | |
| Meta data centres | Erlang chat servers, end-to-end encryption | |
| YouTube | Google data centres | Google infrastructure, VCU transcoding, Bigtable |
| Spotify | Google Cloud | Microservices on Kubernetes, Bigtable, Pub/Sub |
| Snapchat | AWS + Google Cloud | Microservices with an Envoy service mesh |
| Flipkart | Own data centres + Google Cloud (data) | Microservices, Nexus gateway, Aerospike |
| PhonePe | Two active-active data centres | Payments platforms on HBase, Kafka, Druid |
| Paytm | AWS | Thousands of microservices, Paxos databases |
| JioHotstar | AWS (EKS) + multiple CDNs | Microservices behind an Envoy gateway |
| Zomato | AWS | Microservices with Kafka, DynamoDB billing |
| Swiggy | AWS | Microservices with Kafka, DynamoDB, Flink |
| Zerodha | Two data centres + AWS | Self-hosted open source, Go, PostgreSQL |
| Uber | Own DCs + Oracle Cloud + Google Cloud | ~4,000 domain-oriented microservices |
| Not publicly shared | Rest.li microservices, Kafka, FishDB | |
| AWS | PinCompute on EKS, MySQL and TiDB |
All app architecture guides
- Instagram Architecture Explained: Django, PostgreSQL, Cassandra
How Instagram works: Python Django servers, PostgreSQL, Cassandra and Memcache in Meta data centres, its old AWS setup, plus an AWS design you can build. - Netflix Architecture Explained: AWS, Microservices, Open Connect
How Netflix works: Java microservices on AWS, Zuul, EVCache, Cassandra and the Open Connect CDN for video, plus a simple AWS design you can build. - Amazon.com Architecture Explained: Services, DynamoDB and AWS
How Amazon.com works: hundreds of small services, "you build it, you run it" teams, DynamoDB, Aurora and CloudFront on AWS, plus a design you can build. - Prime Video Architecture Explained: AWS Elemental, ECS and Live Sport
How Prime Video works on AWS: Elemental MediaConnect, MediaLive and MediaPackage for live sport, the famous 90% cost-cutting monolith, and a design to build. - Airbnb Architecture Explained: AWS, Rails to Services, Kubernetes
How Airbnb works: a Ruby on Rails monolith split into services, Kubernetes on AWS, MySQL on RDS, photos in S3 and Airflow for data, plus an AWS design to build. - Facebook Architecture Explained: Meta Data Centres, TAO and MySQL
How Facebook works: Meta's own data centres, Hack on HHVM, the TAO social graph over MySQL, Memcache, Twine containers and photo storage, plus an AWS design. - WhatsApp Architecture Explained: Erlang, Encryption, Meta Servers
How WhatsApp works: Erlang chat servers, the Signal protocol for end-to-end encryption, encrypted media blobs, Meta data centres and Rust, plus an AWS design. - YouTube Architecture Explained: Google Data Centres, VCU, Vitess
How YouTube works: Google's data centres, custom VCU transcoding chips, Bigtable and Spanner, Vitess for MySQL and Google Global Cache, plus an AWS design. - Spotify Architecture Explained: Google Cloud, Kubernetes, Bigtable
How Spotify works: thousands of services on Google Cloud, Kubernetes, Bigtable, Pub/Sub, a GCS data lake, Backstage and declarative infra, plus an AWS design. - Snapchat Architecture Explained: Multicloud on AWS and Google Cloud
How Snapchat works: from a Google App Engine monolith to microservices on Kubernetes across AWS and Google Cloud, an Envoy service mesh and Bento ML. - Flipkart Architecture Explained: Data Centres, Google Cloud, Aerospike
How Flipkart works: microservices in its own data centres, the Nexus API gateway on APISIX, Aerospike on Kubernetes, Kafka and Flink, and Google Cloud for data. - PhonePe Architecture Explained: UPI at Scale, HBase, Kafka, Druid
How PhonePe works: active-active data centres, the TStore transaction store on HBase and Elasticsearch, Kafka for billions of events, and Druid analytics. - Paytm Architecture Explained: AWS, Microservices, Paxos Databases
How Paytm works: thousands of microservices across 100+ AWS accounts, EC2 Auto Scaling with Spot, Paxos databases, Redis, Prometheus and Amazon EMR. - JioHotstar Architecture Explained: AWS EKS, Envoy, Multi-CDN Live
How JioHotstar works: microservices on several AWS EKS clusters grouped as data centres, an Envoy API gateway, multi-CDN routing for live cricket and stream checks. - Zomato Architecture Explained: AWS, Microservices, DynamoDB, Kafka
How Zomato works: microservices on AWS with EKS, ECS on Spot, an Envoy mesh, Kafka, DynamoDB billing, Solr search and VictoriaMetrics. - Swiggy Architecture Explained: AWS, Kafka, DynamoDB and CloudFront
How Swiggy works: microservices on AWS with EC2 and Kubernetes, Kong and HAProxy, Kafka as the event backbone, DynamoDB, the Rill Flink platform and CloudFront. - Zerodha Architecture Explained: Self-Hosted Go, PostgreSQL, AWS
How Zerodha works: a small team, self-hosted open-source tools, Go and Python, PostgreSQL, ClickHouse and Kafka, two data centres plus AWS, and simple scaling. - Uber Architecture Explained: 4,000 Services, Kubernetes, Docstore
How Uber works: about 4,000 microservices in domains, Kubernetes across its own data centres, Oracle Cloud and Google Cloud, and Docstore with the CacheFront cache. - LinkedIn Architecture Explained: Kafka, FishDB, Rest.li, LLM Feed
How LinkedIn works: Rest.li microservices, Apache Kafka (created at LinkedIn), the Espresso document store, the Rust-based FishDB feed engine and LLM ranking on GPUs. - Pinterest Architecture Explained: AWS, PinCompute on EKS, MySQL
How Pinterest works: AWS with PinCompute on EKS, MySQL and TiDB, Manas search, Kafka, Flink, Spark and Iceberg for data, and Ray for machine learning.