I am more inclined toward custom solutions using MarkLogic (noSQL) and Apache Hadoop for scaling reasons ;0)