Scheduling parallelizable task in self-organized cloudlet using Hermes

When we have to process a computationally expensive task at our mobile devices, instead of seeking help from remote clouds, we can pool together a set of nearby mobile devices for resource sharing and collaboration through local communications (aka cloudlet). We present Hermes, a system that accelerates parallelizable task completion via disseminating workloads in intermittent local communications. We evaluate Hermes using three Android phones in Nanjing University, and the results show that, Hermes reduces the makespan of the example application (randomized min-cut algorithm) up to 43%.