If you need practical cost efficient time series database, then take a look at VictoriaMetrics. It handles 10 trillions of samples in production in a single-node setup and it can scale to hundreds of storage nodes in the cluster setup.