The author mentioned write performance, but didn't touch much on read performance. Are there any benchmarks you can share in that direction?
Also the author didn't touch much on the space requirements or the node splitting expectations.
Those two areas would be useful to explain in greater detail.
Space requirements can admittedly be a bit higher than some other time-series databases or column stores. Since disk is cheaper than memory, and with easy ways of doing retention, we think its a worthwhile trade-off for people for now. But compression is something we will look to down the road. For our later benchmarks we will include memory/space comparisions to help people evaluate.
As for node splitting, our clustering solution (not yet released) is being worked on quite a bit as well. When its closer to being ready we'll include our thoughts on how to best manage your partitions.
This also plays nicely with configurable data retention policies if desired, e.g., only store raw data for X days, but more aggregated data for longer periods of time.