Popular applications such as email, image/video galleries, and file storage are increasingly being supported by "cloud" platforms in residential, academia and industry communities. The next frontier for these user communities will be to transition "traditional desktops" that have dedicated hardware and software configurations into "virtual desktop clouds" that are accessible via thin clients. The project aim is to develop optimal resource allocation frameworks and performance benchmarking tools that can enable building and managing thin-client based virtual desktop clouds at Internet-scale. Virtual desktop cloud experiments will be conducted under realistic user and system loads by leveraging multiple kinds of GENI resources such as aggregates, user opt-in mechanisms, measurement services and experimenter workflow tools. Project outcomes will help in minimizing costly cloud resource over-commitment, and in avoiding thin client protocol configuration guesswork, while delivering optimum user experience. Further, they will positively impact computer desktop user communities, GENI-like testbeds, and equipment vendors.
The experiments realized using the GENI infrastructure addresses the growing challenges of supporting diverse and flexible network services for billions of mobile devices, one of the most common computing platforms in the future, through the novel support of in-network storage and computation. The design of current network services is mostly limited by the famous end-to- end principle, with the networks acting as a "dumb" pipe, mostly responsible only for transporting bits. So far, the network has often been considered as the sole bottleneck in satisfying application requirements. However, network capabilities have significantly improved with higher speed network links and enhanced switching/forwarding support. Thus the resource bottleneck is gradually shifting towards end-hosts, especially for those resource-constrained mobile devices. Existing network services such as video streaming, although available in the form of managed services, i.e., YouTube, are limited due to restricted end-user control. Such control entails both content management as well as content customization. For example, it is difficult to integrate video sharing with other types of content or restrict access to only a small group of people. With respect to content customization, network services today are simply not designed for mobile users. In this project we address the key challenges due to resource constraints both within the network as well as at the end-hosts. To effectively address the above challenges, our work transformed the way current network services are delivered by exploring the concept ofin-network storage and computation and by exploiting network virtualization. While the idea of in-network storage and computation itself is not new, given that they have been proposed by past work such as the Internet Backplane Protocol and active networks respectively, now the time is truly ripe given the technology trends of cheaper storage and more abundant core network bandwidth, as well as the application demand from resource-limited mobile devices. Such network support creates opportunities to greatly reduce perceived network latency and energy consumption at the end-hosts as well as the link load in both access and core networks. It enables a much more general platform for content delivery for both Internet users and mobile clients, encompassing computation needs required for dynamic content, significantly exceeding today’s CDN support for static content only. Our work have lead to public data repositories and software releases that enable other researchers in the networking and distributed systems areas as well as application developers to benefit from our findings of in-network storage and computation infrastructure support. This project has also supported several graduate and undergraduate research activities.
