In case people are interested, I wrote a post about fault tolerance strategies of data systems like Spark and Flink: https://github.com/marsupialtail/quokka/blo

In case people are interested, I wrote a post about fault tolerance strategies of data systems like Spark and Flink: https://github.com/marsupialtail/quokka/blob/master/blog/fault.md https://github.com/marsupialtail/quokka/blob/master/blog/fau... The key difference here is that these systems don't store data, so fault tolerance means recovering within a query instead of not losing data.