Autonomous Networks and P2P systemsResearchersPh.D. StudentsShort descriptionDistributed aggregationA basic problem in distributed systems is to obtain agreement among a collection of nodes on some joint value, typically characterized as a function on input values, each of which is localized to some node. Examples are consensus, distributed averaging, counting, extremal values, voting, and step functions as threshold crossing indicators. We are particularly interested in scalable solutions where the load at each node or link grows sublinearly with the size of the network, and we are interested in understanding better which functions can be computed in a scalable way, under given accuracy and performance constraints.Existing approaches can be divided into those based on spanning trees, and those based on iterative agreement (gossiping). We are interested in comparing these approaches, and in extending them to address new problems, for instance threshold crossing detection. A topic of particular interest is to study the effect of node and link failures. Besides simulation we are interested in analytic results in areas such as convergence rates, convergence speeds, and behaviour under churn, i.e. dynamic topology changes due to node failures and joins. Our achievements so far are:
Secure aggregationA new research topic is the development of aggregations protocols that are secure in the sense of e.g.:
Modelling of P2P systemsWe study techniques for the specification and verification of peertopeer algorithms, such as the stabilization and maintenance protocols of the Chord overlay network. Approaches include bisimulation proofs of picalculus models, I/O automata and temporal logic (TLA).Projects and Funding
Publications
