Guides

Flipkart Architecture Explained: How Flipkart Handles Big Billion Days

By Ravindra Bagale, AWS & DevOps trainer · LinkedIn · Published: 18 October 2026 · Last updated: 18 October 2026

Short answer: Flipkart runs its shopping services as many microservices in its own data centres (2023), and it moved its big data analytics platform to Google Cloud. All app traffic now enters through Nexus, a central API gateway built on Apache APISIX (2025). Fast data for search, pricing and recommendations lives in Aerospike on Kubernetes, handling about 90 million queries per second across three data centres during sales (2025).

Suppose you buy a phone in the Big Billion Days sale

The sale starts at midnight. You and millions of others open Flipkart and search for a new phone. In simple terms:

1. Your app's request reaches Nexus, Flipkart's central API gateway, which checks login, sessions, bots and rate limits.
2. Nexus routes it to the right microservice: search, product page, pricing, offers or cart.
3. Search results, prices and offers are read from Aerospike clusters in under a millisecond.
4. When sellers change prices or rules, the change is saved in MySQL and copied to Aerospike through Debezium and Kafka (change data capture).
5. Clicks and ad views stream through Apache Flink, which handles about 1 million ad events per second (2026).
6. Order and event data later lands in Flipkart's data platform on Google Cloud for reports.

What architecture style does Flipkart use?

Flipkart today is built on a set of shared platforms used by hundreds of services.

1. One API gateway (2025): before Nexus, each team ran its own gateway features. Nexus, built on open-source APISIX with an etcd-based control plane, targets 99.99% availability and under 10 ms of extra p99 latency, and runs across several data centres.
2. Aerospike as a platform (2025): a team of fewer than ten engineers runs Aerospike for 50+ use cases using the Aerospike Kubernetes Operator.
3. Event-driven updates (2026): the rate card engine uses a transactional outbox: write to MySQL, emit an event, and update Aerospike through Debezium and Kafka.
4. Stream processing (2026): the ads team uses Flink with RocksDB state for exactly-right ad budget counting at 1 million events per second.

Flipkart architecture diagram (public parts)

This diagram shows only what Flipkart or its partners have shared publicly. Press Play to follow a request step by step.

What does Flipkart use? Main technologies

Which cloud provider does Flipkart use?

Flipkart uses a mix:

1. Its shopping services run in Flipkart's own data centres; its Aerospike setup spans three data centres.
2. In 2023 Flipkart moved its data analytics platform (FDP) and content catalog object store to Google Cloud, moving from on-premises Kafka to Pub/Sub (130+ billion messages a day) and Hive to Dataproc.

How it started: MySQL and self-managed clusters

1. Before the central platforms, Flipkart teams ran their own gateways and their own Aerospike clusters on virtual machines.
2. Flipkart's data platform ran on self-managed Hadoop, Hive and Kafka in Chennai and Hyderabad until the 2023 move to Google Cloud.
3. Earlier, Flipkart ran MySQL with high availability and studied replication lag.

How does Flipkart store data?

1. Aerospike: fast key-value data, 170+ clusters (2025).
2. MySQL: source of truth for configuration like rate cards.
3. Google Cloud: about 10 PB batch and 2 PB near-real-time data processed daily.
4. Product image storage and CDN details are not publicly shared.

How does Flipkart monitor its systems?

1. Nexus adds central tracing and observability for every API request (2025).
2. Aerospike clusters send metrics through Prometheus sidecars to Grafana (2025).
3. The full company-wide monitoring stack is not publicly shared.

If we built an app like Flipkart on AWS (illustrative)

Suppose you want to build "BazaarBharat", a shopping site for local sellers. Here is a simple AWS layout.

1. Network: VPC, subnets, route tables, gateways

1. Create one VPC with CIDR 10.0.0.0/16 in the Mumbai region (ap-south-1).
2. Use two Availability Zones (A and B), so one AZ failing does not stop the app.
3. In each AZ make three subnets: public (load balancer, NAT gateway), private app (your code) and private data (databases, cache).
4. Attach one internet gateway (IGW) to the VPC. Public route table: 0.0.0.0/0 → IGW. This route is what makes a subnet public.
5. Private app route table: 0.0.0.0/0 → NAT gateway, so servers can download updates but nobody can connect in from the internet.
6. Private data route table: only the local VPC route. Use VPC endpoints to reach S3, DynamoDB and SQS without going over the internet.

2. Compute: where the code runs

1. ALB or API Gateway at the front.
2. Shop services on EKS in private subnets.
3. Order events through SQS (or MSK for Kafka).

3. Storage and databases

1. S3 + CloudFront for product images.
2. Aurora for orders.
3. DynamoDB for cart and stock counts.
4. ElastiCache for prices and offers, like Flipkart's Aerospike use.

4. Monitoring

1. CloudWatch alarms on latency and errors.
2. X-Ray traces.
3. Load-test before sales.

Request flow diagram

Press Play. The numbered list below the diagram lights up with each step.

Flipkart versus our AWS design