Convergence of Array DBMS and Cellular Automata: A Road Traffic Simulation Case

Array DBMSs manage big N -d arrays, are not yet widely known, but are experiencing an R&D surge due to the rapid growth of array volumes. Cellular automata (CA) operate on a discrete lattice of cells that can be modeled by an N -d array. CA are successfully applied to model fire spread, land cover change, road traffic, and other processes. We made traffic CA simulations possible by array DBMS due to novel components: native UDF language, proactive exec plans, convolution operator, retiling strategy, array versioning, locks, virtual axes, etc. A database approach to CA brings powerful parallelization, data fusion, array processing, and interoperability to name a few. To our best knowledge, our work is the first to run end-to-end CA simulations completely inside array DBMS: we enable array DBMS to simulate the physical world for the first time. Paper homepage: http://sigmod2021.gis.gg/