Browsing by Author "Kaija, Kasperi"

Chord is a distributed hash table solution that makes a set of assumptions about its performance and how that performance is affected when the size of the Chord network increases. This thesis studies those assumptions and the foundation they are based on. The main focus is to study how the Chord protocol performs in practice by utilizing a custom Chord protocol implementation written in Python. The performance is tested by measuring the length of lookup queries over the network and the cost of maintaining the routing invariants. Additionally, the amount of data being exchanged when a new Chord node joins the network and how data has been distributed over network in general is also measured. The tests are repeated using various different networks sizes and states. The measurements are used to formulate models and those models are then used to draw conclusions about the performance assumptions. Statistical measurements of quality are used to estimate the quality of the models. The Ukko high performance cluster is used for running the Chord networks and to execute the tests.