Efficient knowledge representation is a critical enabler of scalable and resource-aware robotic task planning. This paper introduces OmniPlan, a modular planning framework for ROS 2 that supports two interchangeable knowledge representation backends, which are a traditional ROS 2-based knowledge base and a distributed knowledge graph. A systematic empirical comparison of both backends is conducted using a symbolic planner, evaluating wall-clock time, CPU time, energy consumed and CO2 emitted across all stages of the planning pipeline. Experiments demonstrate that the knowledge graph backend performs better than the ROS 2-based knowledge base. These findings substantiate the viability of distributed knowledge graphs as a foundation for resource-efficient and sustainable robotic planning systems.