Massively Parallel Computation via Remote Memory Access

We introduce the Adaptive Massively Parallel Computation (AMPC) model, which is an extension of (MPC) model. At a high level, AMPC model strengthens MPC by storing all messages sent within round in distributed data store. In following round, machines are provided with random read access to store, subject same constraints on total amount communication as Our inspired previous empirical studies graph algorithms [8, 30] using MapReduce and hash table service [17]. This allows us give new much lower complexities compared best-known solutions particular, we show how solve maximal independent set O (1) rounds connectivity/minimum spanning tree (log log m / n both ( δ ) space per machine for constant < 1. memory regime MPC, these problems require poly rounds. results imply that 2-C YCLE conjecture, widely believed hold does not