For the requirements of a production environment, one single Redis instance is far from enough to provide a stable and efficient key-value data service with data redundancy and high availability (HA). Using the replication and persistence of Redis may solve the problem of data redundancy. However, without human intervention, the whole Redis service is not able to be recovered when the master instance is down. While various kinds of solution have been worked out for the HA of Redis, Redis Sentinel, natively supported in Redis since version 2.6, is the most widely used HA architecture. By taking advantage of Sentinel, you can easily build a fault-tolerant Redis service.
As the amounts of data stored in Redis grow rapidly, the processing power and memory capacity of a Redis instance with a large dataset (usually above 16 G) may become a bottleneck for the application. And there are more and more latency or other issues when performing persistence or replication, as the size of datasets...