I would recommend trying to implement a
I would recommend trying to implement a bloom filter for yourself on top of KV and then storing the filter within a key. There's not really any extra special hooks we have within KV that would make that more efficient.
3 Replies
Would be at best as fast and consistent as
get since it'd build on top of it.
Maybe a group of DOs a bit everywhere that can exchange messages to sync the filter.There's distributed bloom filters that let you roll things up so you'd want to shard that way. I don't see why it would be based on get - wouldn't you just keep the filter up to date based on put/delete?
Get would only tell you what's been accessed
Well workers can be evicted, so we need to store the hashes. I was thinking in just a value, so the filter would have a max size of 25MB.