badger is the fastest, for high write loads, that's why i use it

wisckey paper describes a LSM key value store that reduces write amplification for the keys tables by separating them, meaning that changing the values does not incur a penalty on key scanning performance

it also means that you can write indexers that live entirely on the key side without impacting the write performance for new data

as a result, it is far faster for searching as well as not being slowed down by new values being written in those searches

i haven't maintained it properly, and may not work anymore, but there was a GC i built for the event store that was able to scan a 20gb event store to define a set of events to prune/delete and to action it, in ~12 seconds, concurrently with normal operations not being slowed down

Reply to this note

Please Login to reply.

Discussion

No replies yet.