- Recommend the best location in a large grid denoting a spatial region, while each cell resembles some spatial location on the map.
- We consider two main types of facilities (Favorable, Unfavourable) and different types of facilities which are geographically distributed across different grid cells.
- Considering larger size grids, euclidean distance across each pair is computationally heavy and unscalable, we implemented an efficient approach using Voronoi Polygons , and Min Max distance algorithm.
All the implementation are done using Java, and require Java 8+ to run it.
- MPI (Message Passing Interface)
- Spark
- MapReduce
- MASS (Multi Agent Spatial Simulation).
- Finding Concurrency: Task Parallelism → Each facility is computed independently, embracingly parallel).
- Decomposition Pattern: Task Driven Data Decomposition → Each Facility Type determines the task that internally decomposes the spatial grid covering respective facility coordinates.
- Algorithmic Structure: Divide and Conquer → Divide Facility to compute Local Skyline, conquer them to find global Skyline
- Distributed Data Structure: Spatial Grid resembling a multi-dimensional spatial region.
I also created a Demo Rendering Widget to visualize a Spatial Grid.