| The prevalence of distributed computing applications has created an increased necessity to efficiently propagate data between computers. Epidemic algorithms, which propagate updates to a data set in a way similar to the spread of disease in natural populations, utilize a technique which shows significant promise. Previous work has shown that epidemic algorithms can disperse data significantly through a network while alleviating a substantial number of the problems associated with data propagation. Much of this work, however, has been with network topologies which do not accurately mirror the current topology of the Internet. This thesis extends the original experimentation to show that epidemic algorithms applied to a current model of the Internet will maintain the properties of being able to successfully propagate updates through a network.; For this thesis, the behavior and layout of the Internet was simulated by creating an underlying Internet topology which used the Network Simulator 2 and Inet, an autonomous system level Internet topology generator. An overlay network, a logical network that functions independently of the underlying topology, was created on top of the underlying architecture. (Abstract shortened by UMI.)... |
