This GSOC project aims to optimize Nemo execution by performing hierarchical aggregation and controlling its fidelity based on the distance of a geo-distributed data processing environment. The project consists of four main tasks: First, implement and check the correctness of the hierarchical aggregation. Second, evaluate the performance improvement in the point of throughput. Third, add the layer of fidelity control which varies on the different levels of hierarchy. And last, implement a learning-based model which automatically adjusts the hierarchy and fidelity level.


Ji Ho Kang


  • Jeongyoon Eo
  • Won Wook SONG