Enhancing Network Performance and Consumer Experience in Named Data Networking (NDN)

Abstract

Named Data Networking (NDN) is an Information-Centric Networking (ICN) architecture proposed as an alternative to the current Internet infrastructure. Using the publish-subscribe paradigm and a standardised naming scheme, ICN allows the components of a publish-subscribe solution to interact without tight coupling between them.

NDN supports a hierarchical naming scheme that identifies chunks of an object resource such as a service, web page, file, etc., using a content identifier. Content identifiers should be location-free and all identical contents should share the same content identifier. This way, content can be freely distributed and cached within the infrastructure utilising the capacity of routers on the path from a source to a destination, called on-path caching. A hierarchical naming scheme may also enable the functionality of request aggregation at routers. The purpose of this mechanism is to aggregate close-in-time requests for the same content and to propagate only the first of them to a content source(s).

This thesis focuses on the exploration of the on-path caching feature and the request aggregation mechanism of NDN to enhance the network performance and the consumer experience within an Internet Service Provider (ISP)/Autonomous System (AS) network by reducing the intra and inter-network traffic and the content delivery times of consumers. By reducing the intra and inter-network traffic of an ISP/AS network, a reduction of the content delivery times of consumers is also expected.

Popularity and Location-based Caching (PLbC) is a lightweight caching algorithm that utilises the criterion of content popularity and the criterion of the location of routers on delivery paths by incorporating them into the caching decision process to construct a probability. PLbC has been shown to outperform the caching policy proven to perform the best among a number of caching policies evaluated in this thesis, i.e. PC+. PLbC has been shown to yield a 3-6% higher probability of retrieving the content locally from the caches of routers compared to PC+, while caching 11-15% less content within the local ISP/AS network. Yet, PLbC is unable to fully utilise the cache capacity of routers as this approaches the catalog size, i.e. the number of object resources within a network.

Content Sharing-Extended Request Aggregation (CS-ERA) is a mechanism that focuses on content sharing between consumers that request to retrieve the same object resource. For this purpose, CS-ERA exploits the mechanism of request aggregation to bound successive requests to be satisfied within the boundaries of the local ISP/AS network, if a local replica exists. To increase the number of successive requests to be bounded, CS-ERA extends the naming granularity to which request aggregation is ap- plied from a chunk to an object resource. CS-ERA has been shown to outperform the original request aggregation mechanism by aggregating 11-17% more object requests within an ISP/AS network, while satisfying 52-81% more chunks of an object resource locally. The results vary depending on whether on-path caching is enabled or disabled.