Measurement And Estimation Of End-to-end Delay In IP Networks

With the extending scale and increasing complexity of Internet, it is more and more difficult to design and manage networks. Efficiently and exactly depicting Internet running status and network characteristics is the key of network management and maintenance. Delay is an important parameter in network characteristics, whose behavior mostly depends on the performance in one direction. So the measurement and estimation of end-to-end delay is especially important.In this dissertation, a new method named round-trip programming algorithm is proposed for estimating end-to-end delay. It measures the forward and backward delays simultaneously and assumes that the packet which experiences the minimum delay doesn't suffer queuing delay in each direction. The sum of forward and backward delays of different packets is regarded as one packet's'round-trip delay'. Moreover if the packets in two directions which experience least end-to-end delays were sent at different time, this'round-trip delay'would be affected by clock skew. Our algorithm adjusts one packet's sending time to another to avoid the effect of skew. Theoretic analysis show that our algorithm will not increase complexity and practical measurements prove that it performs better compared with existent algorithms.The characteristics of Internet end-to-end delay are analyzed and measurement results show that end-to-end delay time series behaves self-similar and strong non-linear. A functional network model is put forward. Measurement results on three different types of end-to-end delay measurement links indicate that the proposed functional network model could match the characteristics of Internet end-to-end delay.