Dimensioning of the IP-Based UTRAN with Multiple Node Bs for Elastic Traffic Using DiffServ QoS

Abstract

This paper presents dimensioning of an IP-based UMTS Terrestrial Radio Access Network (UTRAN) with multiple Node Bs connected to one RNC, where the applied QoS architecture is based on Differentiated Service (DiffServ) with an integrated Weighted Fair Queue (WFQ) and Strict Priority (SP) scheduling. To provide an appropriate network dimensioning for guaranteeing a desired end-to-end QoS, a general analytical approach is proposed in this paper to dimension the individual links of the Iub interface (between the Node B and the RNC) and furthermore to derive overbooking for the backbone link for elastic traffic. The analytical approach is validated by simulations. Furthermore, based on the proposed analytical approach, important dimensioning rules are summarized.

Joint source and sending rate modeling in adaptive video streaming

“…This clearly evidences the compactness of our model, built on conditionally normal state variables observed during a state transition process that has been properly built: Finally, we have considered the normalized autocorrelation r x ½k defined in (7) and we have evaluated its asymptotical expected value (i.e.for large M). The result of the analysis is reported in (A.4) where we denote by R x ½m9 i,j the ði,jÞ element of the covariance matrix R x ½m, defined in (3) and computed as in (4). Let us point out that, while the index k runs over the ordered set of consecutive frames, the index m in R x ½m runs over the ordered set of consecutive GOPs.…”

“…Let us point out that, while the index k runs over the ordered set of consecutive frames, the index m in R x ½m runs over the ordered set of consecutive GOPs. In (A.4) we recognize that the autocorrelation r x ½k depends in a non trivial way on both intra GOP correlation terms, represented by the elements of the matrix R x ½0, and inter GOP correlation terms, represented by the elements R x ½m,ma0; the latter in turn implicitly depends on the state transition probabilities (see (4)) that characterize the particular model under concern.…”

“…The availability of compact models of multimedia streaming traffic is key in several network design issues. From a networking point of view, models can be exploited in tasks such as dimensioning of core and access network resources [4], derivation of admission control criteria [5], optimization of cross-layer scheduling procedures [6]. A model of streaming sources is useful for characterizing the network traffic.…”