PLATO: Predictive Latency-Aware Total Ordering

PLATO is a predictive total ordering protocol designed for low-latency multicast in datacenters. It predicts out-of order arrival of multicast packets by observing their interarrival times, and delays packets before passing them up to the application only if it believes the packets to have arrived in the wrong order. We show through experimentation on real datacenter-style networks that the inter-arrival time of consecutive packet pairs is an excellent predictor of out-of order delivery. We evaluate an implementation of PLATO on the Emulab testbed, and show that it drives down delivery latencies by more than a factor of 2 compared to the fixed-sequencer protocol.