A new method for detection of super point

A super point is an aggregation point whose flow quantity is larger than a predefined threshold. In recent years, several research papers have offered solutions to the problem in networks with high link speeds. In this paper, we propose an on-line method on detecting super points that guarantees accurate and defines a finite memory requirement. This method consists of three parts: a flow sample & hold process, a bloom filter process, and a removal process. It includes two data structures for aggregation points and the bloom filter. We use the three processes and the two data structures together in our solution. The flow sample & hold process is adopted to sample aggregation points, and the removal process helps to save memory space and to remove some non-super point records. The bloom filter is used as an efficient method to identify a new flow. We provide theoretical analysis and experiments using the NLANR traces.