The invention relates to a distributed monitoring system based on a hadoop cluster and a method thereof. The system comprises a collection agency module, a resource updating module, a network monitoring module, a collection scheduling module, a warning generating module and a hadoop cluster monitoring module, wherein the hadoop cluster monitoring module is connected with the collection agency module, the resource updating module, the network monitoring module, the collection scheduling module and the warning generating module. The method includes the steps of (1) collecting needed detection resources and all index information, (2) storing and caching the collected data, (3) scheduling and updating collected information, (4) timely sampling to generate a data graph needed by a foreground web, (5) monitoring the network state of nodes in the cluster, and (6) determining whether a warning needs to be sent out according to a threshold value and generating warning information if the warning is needed. The distributed monitoring system can serve as an independent cluster monitoring system, a basic index database can be conveniently extended, a current index value can be taken back, a convergence view of indexes can be checked, and the trend of data timing sequences can be observed.