Font Size: a A A

Virtual Computing in Distributed Overlay Networks

Posted on:2012-04-25Degree:Ph.DType:Dissertation
University:University of California, IrvineCandidate:Wang, Richert KaiyouFull Text:PDF
GTID:1458390011952267Subject:Computer Science
Abstract/Summary:
Typical desktop applications, such as word processors, web browsers, e-mail clients, etc., do not fully utilize the computational power of modern personal computers. Personal computing performance has increased at a higher rate than the computational requirements of typical desktop applications. Moreover, with the advent of fast interconnection networks, a large number of underutilized networked resources have become readily available. Overloaded workstations and low-performance mobile devices could harness these idle resources to extend their physical computational constraints and improve execution times of applications.;A novel low-cost high-performance distributed computing environment using existing commodity systems is presented. This dissertation introduces and expands on the Transparent Remote Execution (TREx) environment, which monitors and utilizes idle computational resources within a corporate environment under a fully distributed paradigm. To further extend TREx's local resource manager, the Advanced Quality of Service (AQoS) overlay routing protocol was devised to forward data through less-congested routes on Wide Area Networks (WANs). Thus, a novel virtualization paradigm, Ghost Computing, is capable of distributing coarse-grained Dynamic Link Library (DLL) functions to idle networked resources. Resource demanding applications within this environment can utilize remote idle high-performance computational elements to perform computations that low-performance mobile machines or overloaded workstations cannot handle efficiently. The benefits of TREx and its virtualization capabilities are illustrated with empirical experiments and the performance and accuracy of AQoS are shown using simulation experiments.
Keywords/Search Tags:Computing, Computational, Distributed, Applications
Related items