Dynamic traffic management for interactive cloud services: localising traffic based on network throughput and user mobility

Traffic localisation is an important aspect of traffic management on the Internet. The caching of content and its distribution from localised servers is one mechanism that enables traffic localisation and reduces transit costs for network providers while enhancing service performance. However, personalized, dynamic user content is often un-cacheable due to its nature. As we move towards a world of constantly connected mobile devices, these devices will focus more towards Cloud services as a means of adding capabilities that would otherwise be impossible to have in a small mobile package. This leaves us with the potential problem of having to deal with un-cacheable traffic from real-time interactive Cloud services that causes congestion on a global scale. In this paper we present a solution that considers the scenario of an interactive, personal service running on the Cloud and accessed by a mobile user. We attempt to localise traffic by moving the virtual machine in response to the user's estimated network dwell time. Results gathered from a prototype platform are presented as proof of the concept.