NeTS: Small: Optimal Delivery of Augmented Information Services Over Next-Generation Cloud Networks

University Of Southern California, Los Angeles CA

Investigators

Abstract

Moving data and services from local computers into the cloud has been one of the major trends in information technology over the past years. This trend has upended the computer industry, and provides benefits, for example, in terms of more efficient use of computing resources (which can now be shared by different users) and centralized software updates (which increase security). In a related trend, most information such as video is now stored in the cloud, and delivered to the users on demand, replacing local storage methods such as DVDs and DVRs. Augmented information (AgI) services, an important extension of that trend, combine content delivery and computation, for example, when the original video source data are processed on their way to the destination to create augmented reality, telepresence, real-time surveillance, or industrial automation. This project investigates how such joint communication and computation can be done in the most efficient way in terms of energy, cost, or other criteria that might be important to the consumer or provider. Transmitting information over large distances is usually via multiple "hops", i.e., it goes from the source via several intermediate nodes to the destination. For example, the information could be routed from a source in Los Angeles via Denver, Dallas, Atlanta, to New York. The necessary processing could be done at any of those nodes, or might be distributed among different nodes with different associated costs. Since processing can make the files larger or smaller (e.g., transcoding functions), this also has an impact on the communication cost to the next nodes. This project investigates the fundamental performance limits of this type of network and develops practical algorithms that can do the routing and assignment of processing resources in an efficient way. The goals of the project are to: (i) develop a universal model for AgI services, as a generalization of traditional information services; (ii) formalize the AgI service distribution problem (AgI-SDP) for the quasi-static case, where placement of information and assignment of resources changes only slowly and provide a complexity classification for each major service class and design fast approximation algorithms for the AgI-SDP in its different categories; (iii) develop formulations and algorithms that dynamically adjust the configuration of AgI services in response to unknown changes in cloud network conditions and service demands; and (iv) validate the algorithms by experimentation on testbeds at Bell Labs. The services covered by the AgI model range from network services (e.g., 5G network slices) to automation services (e.g., smart buildings, industrial automation, smart transportation), to augmented experience services (e.g., virtual reality, immersive video), and efficient design of such systems thus has great impact on the economy, ecology (energy savings) and user experience of a wide variety of services. Optimization of such distributed systems also enhance the robustness of computation and communication infrastructure to disruptions from natural disasters and other events. The project includes a strong outreach and education program including having undergraduates and graduate students perform experiments at the project collaborator's state-of-the-art laboratory at Bell Labs. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.