Context. A huge majority of the current Internet traffic is information dissemination. Information-Centric Networking (ICN) is a future networking paradigm that focuses on global level information dissemination. In ICN, the communication is defined in terms of requesting and providing Named Data Objects (NDO). NetInf is a future networking architecture based on Information-Centric Networking principles. Objectives. In this thesis, a global routing solution for ICN has been implemented. The authority part of NDO's name is mapped to a set of routing hints each with a priority value. Multiple NDOs can share the same authority part and thus the first level aggregation is provided. The routing hints are used to forward a request for a NDO towards a suitable copy of the NDO. The second level aggregation is achieved by aggregating high priority routing hints on low priority routing hints. The performance and scalability of the routing implementation are evaluated with respect to global ICN requirements. Furthermore, some of the notable challenges in implementing global ICN routing are identified. Methods. The NetInf global routing solution is implemented by extending NEC's NetInf Router Platform (NNRP). A NetInf testbed is built over the Internet using the extended NNRP implementation. Performance measurements have been taken from the NetInf testbed. The performance measurements have been discussed in detail in terms of routing scalability. Results. The performance measurements show that hop-by-hop transport has significant impact on the overall request forwarding. A notable amount of time is taken for extracting and inserting binary objects such as routing hints at each router. Conclusions. A more suitable hop-by-hop transport mechanism can be evaluated and used with respect to global ICN requirements. The NetInf message structure can be redefined so that binary objects such as routing hints can be transmitted more efficiently. Apart from that, the performance of the global routing implementation appears to be reasonable. As the NetInf global routing solution provides two levels of aggregation, it can be scalable as well.

The current IP based Internet architecture was designed in 70s. The development of new technologies and the evolution of Internet usages make the limitations of this design more visible, especially for the content delivery services. Facing this shortcoming, Van Jacobson and his PARC team proposed Content-Centric Network (CCN) in 2009. The CCN aims to build a content-oriented network, which means the entire networking architecture and all the networking activities are based on the content (content names in particular). The CCN proposal integrates many features such as on-path caching, security, multicast, and native mobility management. This novel proposal has many benefits but it brings also a lot of challenges for current hardware technologies. The transition from IP addresses to content names requires a large memory space to store the content names. However today's fast memory chip cannot meet this requirement. In this thesis I firstly focus on the PIT (Pending Interest Table) element in CCN routers. I propose a distributed PIT system based on the Bloom filter structure to reduce memory requirements and further improve routing performances. The principle of my proposal is that each CCN face manages its own PIT table instead of using a global table read/write lock for a centralized table. This distributed design resolves the information retrieval problem, which is a native shortcoming of Bloom filters. Thus treatment and routing speeds are improved. Thereafter I concentrate on the FIB (Forwarding Information Base) element. In the original CCN design, the FIB is filled by flooding content advertisement. With respect to the huge number of potential content names, this method not only explodes the capacity of FIB tables, but also introduces a high networking traffic. I propose a content-aware CCN forwarding system, which includes a content advertisement publish protocol, a FIB filling algorithm and a downstream forwarding element. In short, the content publish protocol requires that each advertisement is forwarded only towards certain nodes while the downstream forwarding element is a table that is in charge of discovering the other potential content sources. In the third contribution I propose an interface for interconnecting the CCN networking structure with the CDN services. The CDN service has so far been the way to address the content delivery issues of Internet. The key point of interconnecting CCN with CDN service is how to resolve the CDN repository miss-hit problem. The original CCN proposal does not enable efficient interconnection between CDN and CCN. I propose a system, which includes a CDN repository forwarding element and a CDN repository miss awareness mechanism. The former element is in charge of sending the related Interests towards the CDN repositories while the latter one aims to detect the CDN content miss and recover this lost.