Efficient Multiple Aggregations of Stream Data

Recently there has been a great deal of interests in analyzing stream data that can be seen in applications such as network monitoring, web click stream analysis, and sensor networks. Multiple aggregations are regarded as one of the important operations for the high level analysis of stream data as well as business data. However, existing multiple aggregation algorithms for business data are not adequate for stream data because aggregation should be done on a rapidly flowing unsorted data stream, which requires tremendous amount of time and space. We propose an algorithm for efficiently generating user selected aggregation tables from unsorted data stream. For fast aggregation, we use a combination of arrays and AVL trees as temporary storage of aggregation tables. The proposed algorithm can also be used for the cases where aggregation tables are too large to be stored in main memory during aggregation. We showed by experiments that our algorithm is practical.