The document discusses LLAP (Live Long and Process), a new execution engine in Apache Hive 2.0 that enables sub-second analytical queries. LLAP keeps a small subset of frequently accessed data in memory to enable faster query processing times compared to traditional Hive architectures that rely on disk access. It works by running Hive query fragments simultaneously in both YARN containers and long-running daemon processes that cache data in memory. This allows for highly concurrent query execution without specialized YARN configurations. The document provides details on how LLAP is implemented and evaluates its performance benefits based on benchmarks and customer case studies.