A Novel Internet Real-Time Traffic Pattern Detection Technique for Better Pervasive Computing

The Internet follows the power law. For this reason its traffic pattern takes many forms, which change without warning. For example it may change suddenly from LRD (long-range dependence) such as heavy-tailed or self-similar to SRD (short-range dependence) such as Poisson or multifractal. This makes it difficult to run time-critical pervasive applications over the Internet successfully because it is hard to control the response timeliness of the logical TCP (transmission control protocol) channels. The proposed real-time traffic pattern detector (RTPD) technique is generic and detects and identifies LRD and SRD traffic pattern online. If it is implemented as a logical object, then realtime and pervasive applications can use its detected results to self-reconfigure at runtime for better performance that includes shorter service roundtrip time (RTT) and fault tolerance. The RTPD is conceptually the "M 3 RT + R/S + filtration" combination. The M 3 RT (micro mean message response time) tool is the micro implementation of the convergence algorithm (CA), which is an IEPM (Internet end-to-end performance measurement) model with feedback. Alternatively known as the micro CA (MCA), this tool predicts the mean of any waveform quickly and accurately, either on-line or in a postmortem manner with pre-collected traces. A micro IEPM tool operates as an independent object, to be invoked for service anytime and anywhere by message passing. If M 3 RT is inhibited, then RTPD works with the traditional R/S (rescaled adjusted statistics) estimator, but still detects the LRD and SRD patterns on-line. If M 3 RT support is activated, then RTPD works with the enhanced R/S or E-R/S. The proposed RTPD technique differs from other post-mortem approaches (e.g. Hill estimator) because it detects and identifies traffic patterns on the fly. Its contribution to Internet-based pervasive applications is significant because its output enables these applications to reconfigure on the fly and adapt quickly to new operational criteria. The result is better system performance in light of a shorter service roundtrip time (RTT) that makes the client happy