The authors argue that, for communication among the components of a distributed system, latency is as much of a problem as speed and that something can be done about it. Each of the paper’s three major sections (excluding the introduction and the conclusion) contributes to the argument. The authors point out that the latencies of newer communication interfaces (such as Fiber Distributed Data Interface and ATM) have not improved in proportion to their speed. Measurements are presented showing where the delays occur. Next, the design of a low-latency remote procedure call mechanism is described. The purpose is to show that one cannot assume that the bottleneck is necessarily in the higher layers. Again, measurements are presented. Finally, latency-reducing suggestions are made for controller design. Even those who already believe that they know all these points may profit from finding them all in one place.