US Ignite: Collaborative Research: Focus Area 2: Resilient Virtual Path Management for Scalable Data-intensive Computing at Network-Edges

点击翻译按钮获取中文摘要

This project addresses bootstrapping and managing virtual path networks hosting Visual Cloud Computing (VCC) applications for public safety during disaster incidents. This project meets the needs of incident-supporting VCC applications that are essential for first responders to gain rapid visual situational awareness through: (a) scalable processing of imagery/video data collected at incident scenes, and (b) use of thin-client technologies for collaboration. Targeted demonstrations will also be conducted in collaboration with first-responder agencies such as trauma care hospitals (with theater-scale use cases) and a county fire protection district (with regional-scale use cases). Specifically, the project objectives are to: (i) investigate virtual path management algorithms and large-scale simulations with the aim of providing resilient VCC applications with guaranteed performance even in the presence of host/link failures, host mobility and network partitioning; and (ii) demonstrate resilient resource provisioning for such path management algorithms within hybrid fog-cloud infrastructures on experimental testbeds such as GENI and CloudLab.The salient theoretical contribution of this project lays in the design of novel distributed and federated algorithms for the virtual path embedding and computation placement problems, for real-time VCC applications. In particular, the project seeks a solution improvement for constrained path finders schemes, e.g., virtual network embedding or traffic steering, in presence of severe host and link failures, host mobility and network partitioning. The work also advances knowledge of application-aware network design through a scheme that allows fog-cloud compute location selection tradeoffs. In particular, the approach is based on decoupling small/large instances of visual data processing in the context of latency-critical VCC applications at: (a) "theater-scale" for first-responder and incident commander personnel co-ordination at an incident scene, and (b) "regional-scale" for tracking objects of interest across different geographical scales in wide-area motion imagery from airborne platforms.