In the future, users will benefit from a close coupling between their physical realities and a supporting cyber infrastructure. This requires that users be continuously plugged into a virtual world and able to interact with it bi-directionally with quality of service assurances. This vision aligns with the objectives of parallel research projects under the Interactive Digital Media (IDM) subprogram at ADSC that seek to incorporate audio, video, and unstructured Web data to provide a genuine human sixth sense augmented reality and related applications. However, the current ubiquitous access networking system, i.e., the Internet, does not provide Quality of Service (QoS) due to its end-to-end design principle. As part of the Interactive Digital Media (IDM) subprogram, this project aims at enabling QoS support in Internet networking systems so as to support seamless interactions and performance for Telepresence applications.
One fundamental problem is that cellular operators today distinguish communications, in a simplistic manner, as either voice or data access. Whereas voice is supported by predictable TDM virtual circuits, data contents can be arbitrarily classified as non-premium and served through a public Internet without service awareness. On the other hand, IDM data streams are multimedia in nature and rich in semantics. These data streams are of different levels of importance and interact with each other in nontrivial ways, and appropriate levels of QoS must be provided for their transport and processing. The ad hoc approaches currently employed by operators to affect service quality, such as local caching and generation of data and simple service differentiation based on who initiated an access request, do not lead to well-understood impacts on application performance.
In this project, we are developing a rigorous framework that operators can use to admit and assign elastic user tasks to a static pool of distributed and heterogeneous nodal resources in their networks, where the goal is to maximize the aggregate utility of the system, subject to fairness and paid-for minimum QoS constraints for certain applications. In cases where multiple operators need to cooperate, we will consider the underlying issue of economic incentives. We believe that it is timely to re-examine this issue given the emergent trend of 4G wireless, the future prevalence of mobile broadband data access, the technological convergence of communication technologies under IP, the convergence of human societies under globalization, and the emerging marketing opportunities of innovative applications. We are grounding our optimization framework on a model of information-based utility that can capture the performance preferences of real-world users, and will progressively incorporate important real-world constraints including fairness, resource location, and resource ownership into scheduling decisions.