Description
TitleDesign and implementation of real-time cloud-assisted systems
Date Created2017
Other Date2017-10 (degree)
Extent1 online resource (xiv, 106 p. : ill.)
DescriptionThe emergence of various mobile devices (e.g., smart phones, smart wearables, tablets, etc.) has eliminated the location and timing constraints on access to a va- riety of services including social networking, web search, data storage, video streaming, and gaming. Limitations on resources in mobile devices such as battery life, storage, and processing power along with the scalability requirements of such services have mo- tivated companies to rely on cloud computing. Cloud computing has created a new paradigm for mobile applications enabling access to computing and storage resources in various locations on demand. This has drastically transformed the quality of ex- perience (QoE) for the users and increased the popularity of cloud computing based applications. This thesis studies the design and implementation of solutions for sup- port of advanced real-time applications on mobile devices with stringent constraints on responsiveness and throughput using cloud systems. One of the main advantages of cloud computing platforms is to provide a logi- cally centralized but geographically distributed database with on demand access. This characteristic can be leveraged to develop a network layer solution for the problem of maintaining connectivity and global reachability for mobile nodes. In the second chap- ter of this thesis, the design and analysis of a connected vehicle architecture, called FastMF, optimized for the deployment and support of advanced real-time services is examined. Motivated by current in-vehicle needs such as on demand video streaming and safety applications and with a look to a future with cloud-assisted autonomous driving and virtual reality goggles, we identify PHY, MAC, network, and transport layer requirements to support these services. Furthermore, we discuss the issues with current fixed-host/server IP architecture in providing service for mobile nodes and the shortcomings of current cloud systems in terms of reacting slowly to coarse-grained changes in the network. To address these issues, we design a scalable, mobility-centric solution optimized for high dynamicity based on MobilityFirst architecture. Our solution aims at providing a mechanism for multi-hop bidirectional path discovery between vehicular nodes and Internet gateways and enhancing global reachability and seamless connectivity of ve- hicular nodes. To develop vehicular clustering for the purpose of obtaining the best network access and to increase cluster stability, we start with an analytical evaluation of vehicular mobility models followed by link lifetime analysis and description of mo- bility event during a contact between two cars. Our proposed distributed clustering algorithm forms tree-based clusters among nodes with similar mobility attributes. We integrate our proposed clusters into named service layer supported by globally available cloud servers. The distributed global name resolution service (GNRS) implemented in the MobilityFirst architecture along with efficient vehicular clustering enable scalable services with seamless connectivity and global reachability. MobilityFirst services are provided mainly by routers with long-living links and rela- tively consistent connectivity to a network infrastructure. To support dynamic scenarios including connected vehicles, we enable nodes with short-living links and intermittent access to the network infrastructure to act as MobilityFirst routers and extend the MobilityFirst services to multiple-hops. The proposed scheme, assigns unique names to clusters of vehicles. These names are independent of the locations of the clusters and the interfaces. By aggregating vehicles in clusters, assigning unique identifiers to them, and extending GNRS service to ad hoc mode, mobile nodes can benefit from multi-homing and other services which are currently supported by MobilityFirst in a seamless way only for nodes with one-hop access to Internet. The proposed method can be added on top of IP or in an independent way as a network service for highly dynamic scenarios. By large scale simulation of various representative scenarios includ- ing content delivery, web content retrieval, and store-and-forward service, we show the flexibility and efficiency introduced by our design and the improvement in throughput and reduction in delay that FastMF provides. Another main advantage of the cloud computing platform is providing access to enhanced computing resources. This enables providing more computationally complex services such as cloud gaming through thin client devices. In this work, we investigate the design and analysis of a cloud gaming platform with optimized QoE for players. In cloud gaming, the game status updating and frame rendering tasks have migrated to cloud resources. The players receive the rendered game scenes which are streamed to their thin-client devices in real-time over the Internet. One of the main issues at stake is maintaining responsiveness while having a smooth display of the game. What makes this problem more challenging is that the existing solutions for video streaming is not designed for latency-sensitive interactive applications. Furthermore, QoE is a subjective metric and to the best of our knowledge, no quantitative modeling for QoE in interactive systems exists. In this regard, we introduce a novel application layer solution which efficiently adapts itself to variations in the channel and server load. Our developed protocol performs on- line frame selection to dynamically reduce traffic rate and proactively conserve network resources for more fresh frames. This design performs optimization of responsiveness and display smoothness by evaluating the freshness of information perceived by player through video frames. Furthermore, we develop a mathematical analysis of interactive applications and provide an analytical model for interactive real-time systems. Using this model, we develop an objective QoE metric that suits the requirements of gaming systems. This includes the minimization of age of information that the client perceives subject to mitigating frozen screens and maintaining constant display frame rate and smoothness. Through conducting extensive analytical modeling and network simula- tion, we show the applicability of proposed model and protocol in improving the gaming experience for both single-player and multi-player games.
NotePh.D.
NoteIncludes bibliographical references
Noteby Mehrnaz Tavan
Genretheses, ETD doctoral
Languageeng
CollectionSchool of Graduate Studies Electronic Theses and Dissertations
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.
Statistical Profile