DNF: A or B or C

given a set of DNFs, split into conjunctions, assign conjunction ID each conjunction has a pointer to its parent DNF. compute the size of each conjunction.

build K-index table: partition all the conjunctions by its size.

0 -> map[key]postingList 1 -> map[key]postingList 2 -> map[key]postingList

posting list is a sorted list of entry, entry: <conjunction-id, boolean flag(belong or not), score> entries are sorted by conjunction-id, bool-flag ascendingly there is a special post-list Z, for zero-size conjunction

• scoring pruning (Done)
• logging monitoring
• lock for RW indexing
• computation/storage separation: different index-store options: redis, memory
• HA configuration distributed
• http, grpc api
• web UI
• plugin: CDC read