Case Studies > Enabling 2M + Concurrent Users at Twitch with Scale, Simplicity and High Availability

Enabling 2M + Concurrent Users at Twitch with Scale, Simplicity and High Availability

Customer Company Size

Large Corporate

Region

America

Country

United States

Product

Redis Labs Enterprise Cluster
Redis
Cassandra
Elasticsearch
DynamoDB

Tech Stack

Redis
Cassandra
Elasticsearch
DynamoDB
Rails

Implementation Scale

Enterprise-wide Deployment

Impact Metrics

Customer Satisfaction
Productivity Improvements
Digital Expertise

Technology Category

Platform as a Service (PaaS) - Data Management Platforms
Analytics & Modeling - Real Time Analytics
Application Infrastructure & Middleware - Data Exchange & Integration

Applicable Industries

Software
Telecommunications

Applicable Functions

Business Operation
Quality Assurance

Use Cases

Real-Time Location System (RTLS)
Predictive Maintenance
Remote Collaboration

Services

System Integration
Cloud Planning, Design & Implementation Services
Software Design & Engineering Services

About The Customer

Twitch is the world’s leading social video platform and community for gamers, bringing together over 100 million community members. The platform allows users to watch and talk about video games from over 1.7 million broadcasters. Twitch's engineering team is dedicated to providing extremely fast response times, high availability, and reliability for their web application. The platform is architected to handle incredible scale, with over 2 million concurrent viewers at peak times. Twitch uses several open-source technologies to manage their data backend, including Redis, Cassandra, Elasticsearch, and DynamoDB.

The Challenge

Twitch, the world’s leading social video platform and community for gamers, faced the challenge of managing an extremely high volume of concurrent users. With over 100 million community members and up to 2 million concurrent visitors, Twitch needed a robust solution to handle their website-wide chat functionality. The chat rooms often scaled up to 400,000+ users, requiring low latency and high availability to ensure a seamless user experience. Additionally, Twitch's engineering team sought operational simplicity and reliability to focus on delivering the best possible experience to their users.

The Solution

To address their challenges, Twitch chose Redis Labs Enterprise Cluster (RLEC) to power their chat application. Redis was initially selected for its blazing fast performance, operational simplicity, and optimized data structures for rapid, high-scale data processing. As Twitch's infrastructure became more complex, they moved additional functionalities such as token caching and view counting from Cassandra to Redis. Redis Labs provided a managed service with VPC peering to Twitch’s infrastructure, which runs extensively in Amazon Web Services. This setup eliminated the need for Twitch to build operational expertise to manage Redis in production. Redis Labs Enterprise Cluster offered high availability, reliability, and minimal operational overhead, allowing Twitch to focus on enhancing their user experience.

Operational Impact

Redis Labs Enterprise Cluster provided Twitch with zero operational hassle and no maintenance worries.
The solution ensured true high availability with no outages or latency issues, critical for Twitch's high-volume chat application.
Twitch's engineering team experienced low engineering effort required to manage Redis, allowing them to focus on other critical tasks.
The managed service setup with VPC peering to Twitch’s infrastructure simplified operations and reduced the need for in-house operational expertise.
Redis Labs' solution helped Twitch achieve their goal of growing confidence in their infrastructure's ability to handle scale.

Quantitative Benefit

Over 100 million community members and up to 2 million concurrent visitors managed seamlessly.
Chat rooms scaled up to 400,000+ users with low latency and high availability.
Redis Labs Enterprise Cluster eliminated the need for client-side sharding, clustering, and scaling.