查看文章

Aggregation functions, which compute global parameters, such as the sum, minimum or average of local device variables, are needed for many network monitoring and management tasks. As networks grow larger and become more dynamic, it is crucial to compute these functions in a scalable and robust manner. To this end, we have developed GAP (Generic Aggregation Protocol), a novel protocol that computes aggregates of device variables for network management purposes. GAP supports continuous estimation of aggregates in a network where local state variables and the network graph may change. Aggregates are computed in a decentralized way using an aggregation tree. We have performed a functional evaluation of GAP in a simulation environment and have identified configuration choices that potentially allow us to control the performance characteristics of the protocol.