Distributed multilevel storage infrastructure for visual surveillance in intelligent transportation networks

• Main Author: Chih, Wei Ling
• Other Authors: Anwitaman Datta
• Format: Theses and Dissertations
• Language: English
• Published: 2018
• Subjects: DRNTU::Engineering::Computer science and engineering::Computer systems organization::Performance of systems
• Online Access: http://hdl.handle.net/10356/74822
• Institution: Nanyang Technological University

Large volume of data is generated by traffic surveillance devices such as cameras and sensors integrated into an intelligent transportation system (ITS). To deal with the extreme volume and the massively geographically distributed sources of data, we advocate a tiered storage and processing architecture, using edge nodes to augment a centralized backbone, facilitated by network coding and emerging fog computing solutions. For that reason, we design a linear optimization cost model (LO) to capture the network usage (e.g., network bandwidth and storage) with the aid of storage helpers for the network stakeholders. By assuming that the network is static over a window of time, we design an algorithm to construct a deterministic network code that sets up the multicast connections from the source processes to destinations (either storage helpers or data centers) for every feasible solution satisfying the set of graph-theoretic-constraints in the LO model to deal with the networks at small to medium scales. The LO model and the network coding algorithm will be implemented by leveraging OpenFlow, which is a realization of SDN controller. An SDN controller has the global view of the network and operates as a network operating system (e.g., NOX), and thus it is suitable to become the entity to overhaul and make the global decision for ITS. We also emphasize on a hybrid approach of both random and deterministic network coding paradigms to deal with the networks at medium to large scales. We have run several simulations to compare the performance of the ingredients of the hybrid approach by using ns-3 to reconfirm that from the aspects of network throughput and network usage, deterministic network coding performs better than random network coding. We have also compared the performance of the hybrid approach against both network coding paradigms. Under certain network conditions, our simulation results show that the proposed hybrid approach performs better than other approaches.