It's a lot of work to implement the file storage from scratch and there are good options out there (e.g. BigTable which has crazy latency SLAs) If you get the

It's a lot of work to implement the file storage from scratch and there are good options out there (e.g. BigTable which has crazy latency SLAs) If you get the building blocks right it can be fairly easy to build a decent composable OLAP cube. We built the core stack in under 30KLOC and used it at scale for non-trivial stuff. https://medium.com/hstackdotorg/hbasecon-low-latency-olap-with-hbase-d036233cb373 https://medium.com/hstackdotorg/hbasecon-low-latency-olap-wi... There's a lot more good tech than we had 10 years ago and it should be easy (and fun) to roll high-level stacks that have top performance these days. Something that feels like and lightweight like a library but scales out linearly.