In this paper a simple second order scale symmetry is descibed. Because of the application of the central limit theorem to network traffic, which applies whenever the variance of the traffic from individual sources is finite, this scale symmetry is valid asymptotically for a large number of sources independently of the buffer size, the queueing performance parameter under consideration and the source model. Numerical results for on-/off sources show that it gives clear guidance for how system resources must grow in order to cope with increasing traffic, while maintaining or slightly improving queueing performance. Thus, it may be used for dimensioning purposes beyond the reach of sophisticated algorithms. Parallels to equivalent bandwidth approaches are drawn.

Ett generellt recept om hur ett system (kapacitet, kölängden) ska växa om belastningen ökar med en viss faktor presenteras och diskuteras. Prestandaberäkning, QoS, skalering, dimensionering, resursallokering