Sojourns of fractional Brownian motion queues: transient asymptotics

Krzysztof Dȩbicki,Enkelejd Hashorva,Peng Liu
DOI: https://doi.org/10.48550/arXiv.2308.15662
2023-08-30
Abstract:We study the asymptotics of sojourn time of the stationary queueing process $Q(t),t\ge0$ fed by a fractional Brownian motion with Hurst parameter $H\in(0,1)$ above a high threshold $u$. For the Brownian motion case $H=1/2$, we derive the exact asymptotics of \[ P\left(\int_{T_1}^{T_2} 1(Q(t)>u+h(u))d t>x \Big{|}Q(0) >u \right) \] as $u\to\infty$, {where $T_1,T_2, x\geq 0$ and $T_2-T_1>x$}, whereas for all $H\in(0,1)$, we obtain sharp asymptotic approximations of \[ P\left( \frac 1 {v(u)} \int_{[T_2(u),T_3(u)]}1(Q(t)>u+h(u))dt>y \Bigl \lvert \frac 1 {v(u)} \int_{[0,T_1(u)]}1(Q(t)>u)dt>x\right), \quad x,y >0 \] as $u\to\infty$, for appropriately chosen $T_i$'s and $v$. Two regimes of the ratio between $u$ and $h(u)$, that lead to qualitatively different approximations, are considered.
Probability
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the asymptotic behavior of the queue sojourn time probability in a queuing system driven by fractional Brownian motion (fBm) when the system load exceeds a high threshold \(u\). Specifically, the author studies the exact or sharp asymptotic approximations of the sojourn time probability under different conditions, especially in the following two scenarios: 1. **Brownian motion case** (\(H=\frac{1}{2}\)): - For the small - fluctuation case (\(\omega = u+w + o(1)\), where \(w\in\mathbb{R}\)), the sojourn time probability approaches a positive constant. - For the large - fluctuation case (\(\omega=(1 + a)u+o(u)\), where \(a\in(- 1,\infty)\)), the behavior of the sojourn time probability depends on the value of \(a\): - When \(a\in(-1,0)\), the probability approaches 1. - When \(a > 0\), the probability approaches 0 at a speed controlled by \(a\). 2. **General fractional Brownian motion case** (\(H\in(0,1)\)): - The asymptotic approximations of the conditional sojourn time probability are studied, especially for the following two cases: - Small - fluctuation case: \(\omega = u+h(u)\), where \(h(u)\) is relatively small. - Large - fluctuation case: \(\omega=(1 + a)u\), where \(a\neq0\). Through rigorous mathematical analysis and probability theory methods, the paper provides exact or sharp asymptotic expressions for these probabilities, which are of great significance for understanding the long - term behavior of queuing systems driven by fractional Brownian motion. These results are not only helpful for theoretical research, but also can be applied to practical problems in communication networks, traffic flow and other fields.