SUMMARYOver the last few years, autonomic network and service management has emerged as a serious alternative to traditional management approaches. In these systems, distributed entities, called autonomic managers, perform monitoring and control operations in an autonomous and decentralized way. The monitoring consists of providing indicators on the state of the system. Several monitoring solutions have been proposed to enable autonomic managers to obtain a partial or complete knowledge of an indicator through aggregation processes. Such a profusion of solutions raises important questions regarding the choice of an aggregation scheme in a particular operational context and for a particular management information because each class of solution presents different benefits and weaknesses. That is why, in this paper, we present the result of our study of decentralized aggregation schemes for autonomic network and service monitoring. The contribution is twofold: (i) a survey of decentralized aggregation schemes based on a refined taxonomy; and (ii) the results of an evaluation campaign we performed to compare typical aggregation schemes. These results highlight the context, in terms of the managed network behaviour as well as information dynamics, in which each aggregation scheme outperforms the others, thus helping autonomic management system designers in choosing the best scheme for their management purpose.