Moka is a concurrent cache library that I have been developing since 2021. It is used in the backend services of crates.io and sentry.io, for example, to support their high traffic web services.

In this session, I will share my experiences and challenges gained through the development of Moka. I will start with cache specific topics such as improving hit rates. And then I will present some knowledge that can be applied to general concurrent programming, such as how to choose a right data structure that will not spoil concurrency in asynchronous runtimes.

• Requirements of a cache library and how to achieve them.
• Algorithms to increase the hit rate.
• Arc<RwLock<_>>. Simple API for concurrent programming.
• Common technics to increase concurrency.
• Atomic* types, reducing lock contentions, lock-free data structures, etc.
• Challenges and next steps.
• Avoiding bugs related to unsafe Rust and how to debug them, etc.