On distributed mobile edge computing

Mobile Cloud Computing (MCC) has been proposed to offload the workloads of mobile applications from mobile devices to the cloud in order to not only reduce energy consumption of mobile devices but also accelerate the execution of mobile applications. Owing to the long End-to-End (E2E) delay between mobile devices and the cloud, offloading the workloads of many interactive mobile applications to the cloud may not be suitable. That is, these mobile applications require a huge amount of computing resources to process their workloads as well as a low E2E delay between mobile devices and computing resources, which cannot be satisfied by the current MCC technology. In order to reduce the E2E delay, a novel cloudlet network architecture is proposed to bring the computing and storage resources from the remote cloud to the mobile edge. In the cloudlet network, each mobile user is associated with a specific Avatar (i.e., a dedicated Virtual Machine (VM) providing computing and storage resources to its mobile user) in the nearby cloudlet via its associated Base Station (BS). Thus, mobile users can offload their workloads to their Avatars with low E2E delay (i.e., one wireless hop). However, mobile users may roam among BSs in the mobile network, and so the E2E delay between mobile users and their Avatars may become worse if the Avatars remain in their original cloudlets. Thus, Avatar handoff is proposed to migrate an Avatar from one cloudlet into another to reduce the E2E delay between the Avatar and its mobile user. The LatEncy aware Avatar handDoff (LEAD) algorithm is designed to determine the location of each mobile user\u27s Avatar in each time slot in order to minimize the average E2E delay among all the mobile users and their Avatars. The performance of LEAD is demonstrated via extensive simulations. The cloudlet network architecture not only facilitates mobile users in offloading their computational tasks but also empowers Internet of Things (IoT). Popular IoT resources are proposed to be cached in nearby brokers, which are considered as application layer middleware nodes hosted by cloudlets in the cloudlet network, to reduce the energy consumption of servers. In addition, an Energy Aware and latency guaranteed dynamic reSourcE caching (EASE) strategy is proposed to enable each broker to cache suitable popular resources such that the energy consumption from the servers is minimized and the average delay of delivering the contents of the resources to the corresponding clients is guaranteed. The performance of EASE is demonstrated via extensive simulations. The future work comprises two parts. First, caching popular IoT resources in nearby brokers may incur unbalanced traffic loads among brokers, thus increasing the average delay of delivering the contents of the resources. Thus, how to balance the traffic loads among brokers to speed up IoT content delivery process requires further investigation. Second, drone assisted mobile access network architecture will be briefly investigated to accelerate communications between mobile users and their Avatars