Columnar formats allow side stepping of some of the most critical issues that plague time series databases. Each label essentially becomes a column which is then efficient to parse through.