Logistical Networking is one of the end-to-end approaches for globally scalable network storage. It is applied to large-scale distributed network storage system such as web caching, FTP mirroring, Content Distribution Network (CDN), and Data Grid etc. To transfer data to large-scale distributed nodes through Logistical Networking, a scalable and efficient mechanism for one-to-many data transfer is necessary. Multicasting is an efficient mechanism for one-to-many data transfer application. It eliminates redundant packet replication and decouples the size of the receiver set. However deployment of network-layer multicast has not been widely adopted and thus Internet is still incapable of native multicast. Application-layer Multicast approaches do not change the network infrastructure, instead they implement multicast forwarding functionality at end-host. Tree first approach improved scalability of application-layer multicast, however it does not consider global topology information. Constructing data paths without topological information which is congruent with the underlying IP-level topology, the data paths may include unnecessary high latency hops and make network resource usage increased.
In this thesis, we propose a landmark based approach which adds landmark scheme to tree first approach for the scalable application-layer multicast. Measuring network latency to each landmark, we can approximate global topology information and construct topologically-aware data paths. Our results of the performance evaluation indicate that the landmark based approach can reduce average path delay (from source to each member) and total link latency of data paths over the tree first approach, and still offer the scalability of the tree first approach at the same time.
