Dynamic Memory Management for Apache AsterixDB

AsterixDB currently assigns static memory budgets to memory-intensive operators (join, sort, group-by) that never change during execution, leading to memory underutilization and performance degradation under concurrent workloads. This project implements a dynamic memory resource broker for AsterixDB that reallocates memory between operators at runtime based on demand and priority. The broker maintains a reserve pool for immediate grants and a priority queue to track operators approaching memory pressure, allowing for the victimization of low-priority operators when needed. Key deliverables include the resource broker implementation, a memory-adaptive hash join operator, TPC-H benchmark results comparing dynamic vs static allocation, and an extensible interface for future memory-adaptive operators.