Свежая статья от Uber, не быстрое чтиво. Кластер поверх MySQL, кеширующий фронт поверх Redis. Самое интересное - в конце.
* Sharding and cache warming allow it to be scalable and fault tolerant. In fact, one of our largest initial use cases drives over 6M RPS with a 99% cache hit rate with a proven successful failover where all traffic was redirected to the remote region.
* The same use-case would have originally required approximately 60K CPU cores in order to serve 6M RPS from the storage engine directly. With CacheFront we serve approximately 99.9% cache hits with only 3K Redis cores, allowing us to reduce the capacity
* Today CacheFront supports over 40M requests per second across all Docstore instances in production, and the number is growing
https://www.uber.com/blog/how-uber-serves-over-40-million-reads-per-second-using-an-integrated-cache/
* Sharding and cache warming allow it to be scalable and fault tolerant. In fact, one of our largest initial use cases drives over 6M RPS with a 99% cache hit rate with a proven successful failover where all traffic was redirected to the remote region.
* The same use-case would have originally required approximately 60K CPU cores in order to serve 6M RPS from the storage engine directly. With CacheFront we serve approximately 99.9% cache hits with only 3K Redis cores, allowing us to reduce the capacity
* Today CacheFront supports over 40M requests per second across all Docstore instances in production, and the number is growing
https://www.uber.com/blog/how-uber-serves-over-40-million-reads-per-second-using-an-integrated-cache/