It appears that it is common to spin up a single Redis instance on a multi-core machine in order to benchmark it (and usually compare it to other multi-threaded designs). This is wrong, since it is not the way redis was intended to be used.
Real production deployments are probably using redis correctly, but it might be nice to add a more convenient way to spin up a redis cluster from command line, so that people will have an easier way to experiment and benchmark Redis cluster.
maybe something like redis-server --spawn-cluster
TBD.
Comment From: oranagra
we discussed this in the core-team meeting, and it was argued that it's better to improve the existing scripts (utils/create-cluster/) instead of building it into redis-server.
the key point is to have some script that's very easy to use, not meant for production purposes, and just spins up a working cluster with one command line and no need for any multi-step setup process. it can / should be volatile, in the sense that it doesn't create configuration files or persistence, and is mainly meant for quick benchmark or manual tests / experimentation.