12.07.2015 Views

Measurement and Analysis of Skype VoIP Traffic in 3G UMTS Systems

Measurement and Analysis of Skype VoIP Traffic in 3G UMTS Systems

Measurement and Analysis of Skype VoIP Traffic in 3G UMTS Systems

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Table 4: Received packets <strong>in</strong> <strong>UMTS</strong> downl<strong>in</strong>k scenariopayload number mean PIT std. PIT3 B 6 9.46 s 4.49 s21 B 14 1.73 s 3.58 s108 B 817 61.32 ms 16.00 ms112 B 16 3.20 s 45.02 ms4.3 Network Utility FunctionWe <strong>in</strong>vestigate the impact <strong>of</strong> the loss ratio on the NUF, i.e. on the PESQ, by study<strong>in</strong>g thebottleneck LAN scenario described <strong>in</strong> Section 4.1 <strong>and</strong> us<strong>in</strong>g ∆T = 150 ms. By reduc<strong>in</strong>g thespeed <strong>of</strong> the bottleneck l<strong>in</strong>k beyond the need <strong>of</strong> the <strong>Skype</strong> connection, we provoke loss whose<strong>in</strong>tensity is roughly given by the miss<strong>in</strong>g capacity share. Figure 12 shows a scatter plot <strong>of</strong> PESQversus the loss ratio l. Even though the PESQ values are highly vary<strong>in</strong>g even for similar lossratios, a l<strong>in</strong>ear trend can be observed <strong>in</strong> Figure 12 as l <strong>in</strong>creases (as assumed <strong>in</strong> [4]). Simpleregression analysis for l ≤ 0.2 (i.e. a l<strong>in</strong>k speed <strong>of</strong> at least 24 kbps) leads to U m ≃ 1 − 1.5 l.Interest<strong>in</strong>gly enough, the change <strong>in</strong> the relative st<strong>and</strong>ard deviation jumps from σ ≃ 0 to σ ≃ 0.3as soon as loss due to overload occurs, while the damp<strong>in</strong>g <strong>of</strong> the PESQ as a function <strong>of</strong> theoverload <strong>in</strong>tensity is sufficiently described by the m-utility function alone. This behavior isdifferent from the one observed <strong>in</strong> [4], where the sources <strong>of</strong> loss <strong>and</strong> delay variations were<strong>in</strong>dependent <strong>of</strong> each other. On the other h<strong>and</strong>, there is an impact on the PESQ which is due to thefact that a network is used, which is captured by the n-utility function U n = 2.90/3.93 ≃ 0.74(cf. Sect. 3). Thus, the complete NUF <strong>in</strong> the bottleneck LAN case reads U Netw ≃ U n · U m <strong>and</strong>U s = 1.In the <strong>UMTS</strong> scenario (cf. Section 4.2), no significant loss has been observed, which meansU m ≃ 1. However, as seen from the scatter plot shown <strong>in</strong> Figure 13, upl<strong>in</strong>k <strong>and</strong> downl<strong>in</strong>kPESQ3.532.516 kbps24 kbps28 kbps32 kbpsPESQ2.82.62.4upl<strong>in</strong>kdownl<strong>in</strong>k22.21.50 0.2 0.4packet loss ratio20 0.02 0.04 0.06 0.08 0.1σ = (s − s ) / s rcvd sent sentFigure 12: Packet loss probability l vs. PESQ forthe bottleneck LAN scenarioFigure 13: σ = (s rcvd − s sent ) /s sent vs. PESQfor the <strong>UMTS</strong> scenario12

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!