Their basic approach for handling time series data -- partition the inserts by time range -- is one that works for pretty much any relational database. If you can treat the data as immutable which is often the case with time series data, it gets even better. The fact that they have an out-of-the-box solution for it is nice, but I've implemented my own for an application in which a portion of the workload was time series data and a separate portion was not. It's all about mechanical sympathy and understanding the performance characteristics of your chosen datastore.