A model of parallel computation in which one keeps track of the total work and depth of a computation without worrying about how it maps onto a machine.