Abstract :
A growing number of applications have become reliant or can benefit from monitoring data streams. Data streams are potentially unbounded in size, hence, Data Stream Man- agement Systems generally maintain a "sliding window" containing the N most recent elements. In an environment where the number of stream sources can vary, the amount of storage available to hold the sliding window can reduce dramatically. However, it has already been noted that as data becomes older their relevance tends to diminish un- til they are ultimately discarded from the sliding window. Based on this assumption, we propose to "wound" older data elements by relaxing their storage requirements as an effort constantly free enough space to keep pace with accu- rate representation of incoming elements in a process that we call aging. We propose two incremental quantization techniques that enable aging in an efficient manner. We will show that, by relaxing storage utilization of the summary created by our quantizers, the older data elements are not rendered useless. In fact, we will show that their accuracy is only lessened by a sustainable amount.
