Finite Length Coding in Edge Computing Scenarios

Abstract

Future cellular networks are expected to have resources for data processing at the edge of the network, providing the benefits of cloud computing without long routing delays. Such computing resources may for example be used in a control scenario, where wireless sensors upload data to a controller running at the edge of the mobile network. The controller computes an actuation command that is then sent back on the downlink to a wireless device. The delay in such a system not only depends on the processing time of the controller, but also on the uplink and downlink channels where fading and packet losses can result in a queuing delay. In this work, we present a probabilistic upper bound on the total delay of such systems when the channels are subject to Nakagami-m fading. Our method takes the effect of channel coding at finite blocklength in the uplink and downlink channels into account. Using several series expansions, the probabilistic delay bounds can be computed analytically, providing guidelines for resource allocation without the need for extensive simulations.