An Information-Centric Approach to Quality of Service in a Highly Dynamic Edge Environment
Citation:
MCCARTHY, JESSICA, An Information-Centric Approach to Quality of Service in a Highly Dynamic Edge Environment, Trinity College Dublin.School of Computer Science & Statistics, 2019Download Item:
Abstract:
There is a fundamental paradigm shift in how today's communications networks are being used to deliver data. This has emerged from the proliferation of intelligent Internet of Things devices and corresponding volumes of data being generated by these devices. This data explosion is not just an increase in the sheer volume of the data. Data flows are dramatically different, with machines creating and consuming the data like never before. In addition, connectivity to the network is often heterogeneous and opportunistic. The Internet Protocol (IP) is the universally-accepted networking protocol that has served us very well, transforming networks into the age of information. However, IP was designed with the sole purpose of connecting two machines over a fixed wired network. The requirements for the network when IP was designed were vastly different from the requirements we now have. Demand for heterogeneous content is rising sharply and more efficient content distribution is required to manage the corresponding traffic, and return requested content with acceptable Quality of Service (QoS). QoS provisioning and fulfilment is dependent on content recognition, which is not supported natively in IP host-centric networks. In IP,overlay mechanisms are used to capture content information to be used for QoS fulfilment,but this approach is inefficient given current network requirements. Information-Centric Networking (ICN) provides content-based delivery, but QoS concerns are not sufficiently addressed. The vehicular application domain is used in this work as an exemplar of one with complex networking requirements. Modern vehicles include significant technology that could be exploited to improve safety and road efficiency, but making use of these technologies requires time-sensitive information delivery between vehicles. The capabilities of the network are of critical importance to provide the required quality of service of data delivery. This thesis presents a QoS-aware Information-Centric Network for vehicular applications. In particular, the work extends the Named-Data Network (NDN), a variant of ICN, focusing on data delivery deadline awareness. The new algorithms classify the priority of requests, with associated QoS requirements by encoding QoS information into the interest request packets and corresponding data reply packets, and extending the routing algorithm to use multihop forwarding to efficiently request and receive the requested content without prebuilding routes. The work also explores extending network traffic control mechanisms at the data link layer, to assess the potential impact on network congestion management. Evaluations have been carried out using extensive simulation, in particular using a combination of the ndnSIM and SUMO simulators. Deadline aware success rate and packet success rate are both measured under different network densities, vehicle speeds, proportions of vehicles in the environment acting as content producers, and experiment durations. The QoS-aware ICN algorithms are assessed against four baselines: UDP IP (with both DSRC and WiFi communication channels), and basic NDN (DSRC and WiFi). The results of this study demonstrate that the QoS-aware approach generally achieves higher success rates at delivering different packet priority types within their deadlines, relative to the baselines. In addition, these success rates are achieved with fewer packet retransmissions. However, adding QoS-awareness to network traffic control at the data link layer is less impactful.
Author's Homepage:
https://tcdlocalportal.tcd.ie/pls/EnterApex/f?p=800:71:0::::P71_USERNAME:JMCCART7Description:
APPROVED
Author: MCCARTHY, JESSICA
Advisor:
CLARKE, SIOBHANQualification name:
Doctor of Philosophy (Ph.D.)Publisher:
Trinity College Dublin. School of Computer Science & Statistics. Discipline of Computer ScienceType of material:
ThesisCollections
Availability:
Full text availableMetadata
Show full item recordLicences: