[PDF]

Abstract

To achieve immersion, Mixed Reality (MR) applications need to strict QoS requirements. To do so on light-weight user devices, MR applications may need to offload some processing tasks onto more powerful compute nodes. While Mobile Edge Computing (MEC) provides such nodes in close network proximity, MEC nodes may be oversubscribed. To ensure fairness, MEC systems enforce runtime limits on compute tasks. We propose MicroLambda - a framework to partition offloaded computation over multiple lambda invocations to fit within task runtimes. We also explore avenues for joint scheduling of computation placement and network scheduling within virtualized 5G network slices.