Resource dimensioning through buffer sampling

M.R.H. Mandjes, R. van de Meent

15 Citations (Scopus)

Abstract

Link dimensioning, i.e., selecting a (minimal) link capacity such that the users’ performance requirements are met, is a crucial component of network design. It requires insight into the interrelationship among the traffic offered (in terms of the mean offered load , but also its fluctuation around the mean, i.e., ‘burstiness’), the envisioned performance level, and the capacity needed. We first derive, for different performance criteria, theoretical dimensioning formulas that estimate the required capacity $c$ as a function of the input traffic and the performance target. For the special case of Gaussian input traffic, these formulas reduce to $c = M + \alpha V$, where directly relates to the performance requirement (as agreed upon in a service level agreement) and $V$ reflects the burstiness (at the timescale of interest). We also observe that Gaussianity applies for virtually all realistic scenarios; notably, already for a relatively low aggregation level, the Gaussianity assumption is justified. As estimating $M$ is relatively straightforward, the remaining open issue concerns the estimation of $V$. We argue that particularly if corresponds to small time-scales, it may be inaccurate to estimate it directly from the traffic traces. Therefore, we propose an indirect method that samples the buffer content, estimates the buffer content distribution, and ‘inverts’ this to the variance. We validate the inversion through extensive numerical experiments (using a sizeable collection of traffic traces from various representative locations); the resulting estimate of $V$ is then inserted in the dimensioning formula. These experiments show that both the inversion and the dimensioning formula are remarkably accurate.
Original language Undefined 10.1109/TNET.2008.2009989 1631-1644 14 IEEE/ACM transactions on networking 17 5 https://doi.org/10.1109/TNET.2008.2009989 Published - 2009

Keywords

• Quality of Service
• Large deviations
• EWI-17733
• METIS-266456
• Network dimensioning
• Buffer sampling
• Inversion
• Gaussian traffic
• IR-70529