Redis cluster performance on a single machine vs KeyDB multithreading

A few interesting points taken from this article: https://docs.keydb.dev/blog/2019/10/28/blog-post/

A single Redis node may not be taking advantage of the machine it is running on. Redis is single threaded and benchmarking shows that until you reach a 7 node Redis cluster will the throughput plateau. Its hard to find information on what cluster performance looks like on a single machine as it reaches its limitations. It is understood that sharding to separate machines gets performance gains in a linear fashion but the results below on a single machine are interesting.

Ideally sharding should only be required when you need additional machines to handle your traffic and demands. KeyDB is a fork of Redis that is multithreaded and does just that. This enables a single node to take full advantage of a machines resources without sharding.

The below chart is of a single multithreaded KeyDB node where more threads are allocated.

If you want a deeper dive take a look at the originating article that discusses sharding and cluster limitations, benchmarking, and multithreading uses. You can also check out the KeyDB open source project on Github.