Customer Story

Serving millions of concurrent users in real-time with Aerospike

About the company

Dream11

Dream 11 Logo

Dream 11 is the world’s largest fantasy sports platform, with millions of users engaged in contests related to 12 different types of sports, covering 12,000 real life sporting events a year. With Dream11, Indian sports fans can enter contests to showcase their knowledge of cricket, football, kabaddi, basketball, hockey, volleyball, handball, rugby, futsal, American football & baseball. From 2021 to 2023, Dream11’s user base grew from +100M to +200M users, and 35TB data a day. At peak usage, the platform currently handles 10.56M concurrent users, up from 5.5M concurrent users in 2021.

Challenge

Scaling for high user engagement in real-time

In the minutes leading up to a sporting event, tens of thousands of Dream11 users engage with the platform to join contests, create teams, modify teams, and interact with each other. The activity causes Dream11 to experience a huge spike in traffic, needing to handle well over 308M requests per second at the edge. Real-time performance, extreme scalability, and high availability are key during these moments because one minute of downtime costs Dream11 $1M in lost revenue.

In 2020, Dream11 decided they needed a solution that could scale seamlessly to meet their business needs and maintain the trust of their customers. The legacy infrastructure, which used Amazon RDS as the persistence layer, suffered from numerous challenges including:

Unable to handle increasing user and data volumes

With upwards of 308M requests per second at the edge and 10.56 concurrent users at peak times, the platform was suffering from issues such as poor pagination

Inability to manage periodic data backup

The system was becoming unable to handle periodic archival management of huge volumes of data

Poor performance with mixed workloads

When the platform experienced high write transactions per second, the system experienced bottlenecks in read performance

Inflated costs with scale

As Dream11 tried to scale and grow, costs inflated and became unmanageable
Solution

A modern, reliable, high performance solution with a smaller footprint

With Aerospike, Dream11 modernized their infrastructure to achieve high performance and scalability with a smaller footprint and lower costs. Dream11 uses Aerospike extensively in many of their microservices applications, several of which are critical to the user experience.

In addition to providing reliable high performance, Aerospike has also proven to be extremely straightforward to manage and eases day-to-day operations for the Dream11 team. They are planning to increase their use of Aerospike as they continue to build new services and offerings to their users.

Dream11 uses Aerospike to achieve:

Horizontally scalable architecture

Required a solution that could scale out elastically to higher traffic during busy months and then back to normal traffic at other times

High availability and strong consistency

99.99% uptime with zero outages and support for strong consistency

Fault tolerance and durability

Highly stable application with no data loss

Smaller footprint

Aerospike replaced both the persistence and caching layer
Results

Seamless and smooth user experience, with room to expand

With Aerospike, Dream11 is now able to reliably provide a seamless and smooth user experience, even at peak usage levels. Beyond solving Dream11’s immediate needs, Aerospiike also allows Dream11 to make continuous improvements to performance as well as additionally reduce footprint by taking advantage of the newest hardware and processor offerings, such as Graviton.

1.3M reads per second

Improved scalability enables Dream11 to process records at a rate of 1.3M reads per second and over 0.5M writes per second

<15 ms latency

Transactions are processed at a low latency of under 15 milliseconds

99.99% availability, strong consistency

Reliable 99.99% uptime with strongly consistent reads and writes

Automatic sharding

Operational simplicity utilizing automatic sharding

Surpass limits with Aerospike