Operating 170 clusters for hyperscale e-commerce

Founded in 2007, Flipkart has grown into one of the world’s largest e-commerce platforms, serving more than 200 million users across electronics, fashion, groceries, and household products. Millions of customer interactions occur across Flipkart’s platform every minute. Services such as search, product listings, personalized recommendations, and checkout rely on real-time access to large volumes of data.

Demand can spike dramatically during major sales events like Big Billion Days, one of the largest online retail events in India. Maintaining fast, reliable responses across these services requires a highly scalable real-time data platform.

In 2018, Walmart acquired a 77% stake in Flipkart in a $16 billion transaction.

Operating a massive real-time retail platform

Flipkart runs a highly interactive e-commerce platform where every customer interaction triggers multiple real-time data operations.

Services such as product discovery, recommendations, listings, and search ranking rely on real-time data to respond in milliseconds. At the same time, the platform must handle enormous transaction volumes and unpredictable demand spikes during major sales events.

To maintain a consistent user experience under these conditions, Flipkart needed a database platform capable of supporting:

• Extremely high transaction volumes

• Real-time access across multiple application services

• Sudden and unpredictable traffic spikes

Fragmented Aerospike deployments created operational complexity

To support these workloads, different engineering teams across Flipkart adopted Aerospike independently for various services.

Over time, this resulted in more than 40 Aerospike Community Edition clusters running on virtual machines, each managed separately by individual teams.

This decentralized approach created several challenges.

Scattered deployments: VM-based clusters were configured differently across teams, making it difficult to standardize deployments and operational practices.

Resource inefficiency: Virtualization overhead and uneven workload distribution made it difficult to fully utilize hardware resources.

Operational complexity: Managing dozens of independent clusters required significant manual effort for scaling, configuration changes, and infrastructure maintenance.

As Flipkart’s platform continued to grow, the company needed a way to centralize Aerospike operations and manage the database as a shared service across the organization.

A centralized Aerospike platform powered by Kubernetes

Flipkart created a dedicated platform team to operate Aerospike as a shared service across the company.

The team consolidated existing deployments, upgraded to Aerospike Enterprise Edition, and adopted Kubernetes to automate cluster management.

Using the Aerospike Kubernetes Operator (AKO), Flipkart standardized how Aerospike clusters are deployed, scaled, and maintained across the organization.

Centralized database operations: Aerospike is now managed as a shared platform service, improving consistency and simplifying operations across teams.

Automated cluster management: AKO automates deployment, scaling, and lifecycle management, reducing manual effort and operational overhead.

Efficient infrastructure usage: Running Aerospike on bare metal through Kubernetes eliminates virtualization overhead while maintaining flexible orchestration.

Operating Aerospike at massive scale

With Kubernetes and the Aerospike Kubernetes Operator, Flipkart transformed Aerospike from a collection of independent deployments into a centralized data platform.

Today, the company operates more than 170 Aerospike clusters, supporting real-time services across the platform, including search, product listings, recommendations, and customer insights.

Standardized cluster management allows Flipkart’s platform engineering team to automate provisioning, scaling, and maintenance across its entire Aerospike environment.

This architecture enables teams across the organization to rely on Aerospike as a shared platform while maintaining consistent performance for real-time workloads.

"Since the time we went live with the Kubernetes infra, we have seen 100% availability with the operator. There hasn't been any incident where we had a production issue because of the operator. It has been working very well for us."

Sharath B.P.

FlipkartSoftware Architect

Delivering extreme scale with operational reliability

Extreme transaction scale

During its Big Billion Days sale event, Flipkart processed 95 million transactions per second, one of the largest transaction volumes seen in global e-commerce.

Operational reliability

The Aerospike Kubernetes Operator has delivered 100% availability with no production incidents caused by the operator.

Centralized platform operations

Flipkart now operates more than 170 Aerospike clusters, enabling teams across the organization to consume Aerospike as a shared platform service.