Miscellaneous

Gonçalves R, Almeida PS, Moreno CB, Fonte V, Preguiça N.  2012.  Dotted Version Vectors: Efficient Causality Tracking for Distributed Key-Value Stores. DAIS - IFIP International Conference on Distributed Applications and Interoperable Systems.. poster_dais.pdf
Jesus P, Moreno CB, Almeida PS.  2007.  A Study on Aggregation by Averaging Algorithms. Abstractjesus-eurosys2007b.pdf

With the advent of multihop ad-hoc networks, sensor networks and large scale overlay networks, there is a demand for tools that can abstract meaningful system properties from given assemblies of nodes. Distributed aggregation algorithms allow the evaluation of properties such as: network size; total storage capacity; average load; and majorities. A useful class of aggregation algorithms is based on averaging techniques. Such algorithms start from a set of input values spread across the nodes, and iteratively average the values in neighbour nodes that are able to communicate. Eventually all nodes will converge to the same value and can estimate some useful metric. For instance, if one node starts with input 1 and all other nodes with input 0, eventually all nodes will end up with the same value and the aggregate property network size can be estimated by the fraction 1.