In-database connected component analysis
share
We describe a Big Data-practical, SQL-implementable algorithm for efficiently determining connected components for graph data stored in a Massively Parallel Processing (MPP) relational database. The algorithm described is a linear-space, randomised algorithm, always terminating with the correct answer but subject to a stochastic running time, such that for any ϵ>0 and any input graph G=〈 V, E 〉 the algorithm terminates after O( |V|) SQL queries with probability of at least 1-ϵ, which we show empirically to translate to a quasi-linear runtime in practice.
READ FULL TEXT