A comparative evaluation of mobile agent performance for network management

Despite the strategic and software engineering benefits mobile agents (MAs) brings to network management, their performance is still a controversial issue. A number of quantitative analyses and experiments on mobile agent performance have been reported in the last few years. Among the claims in these studies, there are some obvious controversies. In an effort to determine the existence of contradiction and explore the cause for disagreement, this paper compares a number of evaluative studies of MA performance in the network management domain. Their experiments and analytical models are briefly described, and their findings and conclusions are highlighted for effective contrasting. With direct inference from this comparative survey, we suggest that many factors must be carefully taken into account when evaluating the MA network management paradigm. These key factors include the size of the network, the specific management tasks the MA is to perform, the initial MA size, data compression, the transfer mode, the specific platform adopted, etc. Our careful examination reveals that most of the disagreement is caused by difference in the above measurement factors. Ample evidence from the reviewed studies demonstrated that MAs are not actually efficient enough for real-time polling and collecting large amount of data, but we have enough proof that MAs demonstrate considerable robustness and performance in performing complex local computing, data filtering and updating. One contradiction in the body of literature is identified when the network has limited size. More carefully planed study is needed to solve this obvious contradiction.