i don't see the point in using a relational database for timeseries, i am never going to do an SQL join on these series.
i'm using arctic https://github.com/manahl/arctic as a timeseries store and am very impressed (speed and speaking pandas DataFrames natively), also doing csv backups though :)