Towards a Lightweight Mobile Cloud

A typical cloud-computing environment is characterised by the networking of tens of thousands of servers. Once deployed, such datacenters experience a relatively stable hardware and networking environment. Smartphones offer considerable processing, memory, storage and sensing resources, and are handheld, pocket-sized devices that users generally carry with them at all times. The vision for cloud computing could foster more flexible models, whereby general users, including mobile ones, could participate in the cloud as both providers and consumers of resources. If the cloud is to computing what the Internet is for data, the opportunity for mobile computers to participate in service provision should be investigated. The motivation for this dissertation is threefold. First, it explores the forces that resulted in the emergence of cloud computing. Second, the implications of mobile server nodes within a cloud environment are considered. Finally, the dissertation focuses on implementing a mobile storage cloud. Such a system could support applications including disseminating critical information in a disaster zone, mobile distributed social networking, and sensor applications such as collecting data on traffic movements. The background analysis indicates several developments that fostered the emergence of cloud computing. During the 1960s virtualisation was developed for allocating mainframe resources, but later waned due to the commoditisation of hardware, only to re-emerge for allocating computing resources within datacenters. Improvements throughout the 1980s and 1990s reduced commodity hardware and networking costs, stimulating migration to horizontally scalable datacenters. Finally, automating resource provisioning via online portals enabled the flexibility and efficiency of federated computing to become widely accessible. Evaluating the implications of mobile server nodes highlights numerous constraints for implementing a mobile cloud. Most importantly it becomes evident that the traditional cloud architecture will not suffice. A smartphone will not currently support a hypervisor or multiple virtual machines. Mobile devices are not designed to act as always-on servers, as they are battery powered, with limited computing resources, and experience low and intermittent network bandwidth. A mobile cloud will therefore only support applications that are designed for such an operating environment. Service availability takes on a different meaning within a mobile cloud as nodes regularly arrive and churn off the network.