I recently looked into more details of https://github.com/stormprocessor/storm-state
and I find it pretty interesting.
I am considering for my bolts to keep some state in memory and snapshot it to a cache or a distributed file system rather than using an external DB or KV store, and storm-state seems to do that very efficiently.
Is there any reason why very few people seem to use it (according to the number of related topic in the mailing list)?
Also the last commit is from 1 year ago, what is the future for this library? Especially, what is going to happen to it when Storm moves to the Apache process in the next release? Does it work with the newest Hadoop API?