The Canva Session Revocation Makeover: A Case Study in Scalability
Canva, the popular design platform, has recently unveiled a significant overhaul of its session revocation architecture, and it's a fascinating case study in managing scalability challenges. The new system is designed to handle the authentication of hundreds of millions of active sessions, and it does so with remarkable efficiency.
The Challenge of Scale
As Canva's user base grew, the traditional approach of storing session data in a networked database became a bottleneck. With hundreds of gateway instances, the system was burdened with an overwhelming number of database lookups, leading to performance issues and increased infrastructure costs. This is a common problem for platforms experiencing rapid growth, and it's a testament to Canva's engineering team that they tackled this challenge head-on.
Amazon S3 to the Rescue
Canva's solution was to leverage Amazon S3, a simple yet powerful move. By storing revocation data in S3 as compact, immutable records, they've created a highly scalable and efficient system. This approach reduces the need for frequent database interactions, which is crucial for maintaining performance as the number of users skyrockets. What's particularly clever is the use of S3 objects with a 30-minute window, allowing gateways to fetch only the data they need, when they need it. This design choice significantly reduces the memory footprint and improves deployment speed.
The Power of In-Memory Indexes
The use of in-memory indexes is a game-changer. Gateways can now authenticate requests locally, without the latency of database lookups. This not only speeds up the process but also reduces the risk of single points of failure. In my opinion, this is a brilliant strategy for ensuring high availability and performance, especially in a distributed environment. The fact that Canva can now process over 2,000 revocations per second is a testament to the effectiveness of this design.
A Refreshing Take on Refresh Tokens
One interesting aspect of this redesign is Canva's response to the suggestion of using refresh tokens. While it's a common approach, Canva engineers argue that it would increase database load due to frequent token refreshes. This is a nuanced perspective, as it highlights the trade-offs between different authentication strategies. In their case, keeping revocation data in memory provides better overall performance, demonstrating that there's no one-size-fits-all solution in system design.
Implications and Takeaways
Canva's new session revocation architecture offers several valuable lessons for engineers facing similar scalability challenges. Firstly, it emphasizes the importance of choosing the right tools for the job, like leveraging S3 for its durability and simplicity. Secondly, it showcases the power of optimizing data storage and retrieval, which can significantly impact system performance. Lastly, it reminds us that scalability is about finding the right balance, often through innovative design choices. Personally, I find this a great example of how a thoughtful redesign can not only solve immediate problems but also set the stage for future growth.
