Pregled bibliografske jedinice broj: 152299
Statistical Properties of Low-rate Internet Traffic and Application to Traffic Engineering
Statistical Properties of Low-rate Internet Traffic and Application to Traffic Engineering // Proceedings of the 27th International Convention - MIPRO / Golubic, Stjepan ; Mikac, Branko ; Hudek, Vlasta (ur.).
Opatija: Hrvatska udruga za informacijsku i komunikacijsku tehnologiju, elektroniku i mikroelektroniku - MIPRO, 2004. str. 100-105 (predavanje, međunarodna recenzija, cjeloviti rad (in extenso), znanstveni)
CROSBI ID: 152299 Za ispravke kontaktirajte CROSBI podršku putem web obrasca
Naslov
Statistical Properties of Low-rate Internet Traffic and Application to Traffic Engineering
Autori
Vrdoljak, Zdenko ; Gracin, Josip ; Pilipović, Smiljan
Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni
Izvornik
Proceedings of the 27th International Convention - MIPRO
/ Golubic, Stjepan ; Mikac, Branko ; Hudek, Vlasta - Opatija : Hrvatska udruga za informacijsku i komunikacijsku tehnologiju, elektroniku i mikroelektroniku - MIPRO, 2004, 100-105
Skup
MIPRO 2004 27th International Convention
Mjesto i datum
Opatija, Hrvatska, 24.05.2004. - 28.05.2004
Vrsta sudjelovanja
Predavanje
Vrsta recenzije
Međunarodna recenzija
Ključne riječi
traffic engineering; traffic statistics
Sažetak
It is widely believed that non-guaranteed (i.e. best-effort) Internet traffic has a slowly decaying autocorrelation function while its remaining statistical parameters correspond to long range dependent stochastical processes, such as self-similar processes. However, after performing extensive analysis of certain publicly real network traces, we found that such an assumption is not valid for low-rate best-effort Internet traffic with mean rates less than 40MB/s. In addition, we found that heavily-tailed distributions do not describe the rate distribution efficiently. Other subexponential distributions, such as the Weibull and Log-normal distributions, seem to capture the real distribution much more accurately. Due to their finite means and variances, it is very easy to fit the mentioned distributions into the empirical distribution based only on values of the first two moments. In conclusion we suggest a way in which to apply these principles in order to create a simple capacity planning procedure.
Izvorni jezik
Engleski
Znanstvena područja
Elektrotehnika