Date Added: Jan 2011
Many applications, such as data-mining and network traffic monitoring, search vast amounts of data to uncover general trends. In many cases, the sheer volume of data to be examined can preclude multi-pass approaches. Some data, such as packets in a network, are most naturally abstracted as streams, structures in which access is required to be serial (As opposed to random access) and for which the length is potentially infinite. This constrains one to consider data characterization algorithms that only examine data in a serial manner and store a small window of the entire stream at any given instant.