Why it matters
LLAP is Hive's interactive path. Understanding shapes low-latency Hive.
Advertisement
The architecture
LLAP daemons: long-running JVMs on cluster nodes.
In-memory cache: shared across queries.
Query fragments run in daemons.
Advertisement
How it works end to end
Cache: columnar cache with LRU eviction. Shared across queries + users.
Fault tolerance: query re-run on daemon death.
Integration: Tez uses LLAP for supported queries; falls back to containers.