Adaptive bandwidth control for efficient aggregate QoS provisioning

Abstract

Abslmci-This paper proposes an adaptive bandwidth control algorithm that efficiently provides aggregate loss guarantee to resolve the problem of inefficient bandwidth allocation doe to incomplete, inaccurate traffic descriptors supplied by users. Because the control attempts to alloeate the bandwidth only just enough to meet the QoS requirement, the amount of baudwidth saving compared to static aUocation can be substantial. Another distinct advantage of our control algorithm i s that no a priori information on the traffic characteristics of the aggregate is required. From the simnlation study, the proposed control can maintain the packet loss QoS while attaining very high utilization, and is robust against different system configurations and controller parameters.

The QoSbox: Quantitative service differentiation in BSD routers

“…From the algorithmic perspective, the notion of using a dynamic rate allocation to react to traffic arrivals is not new, and was for instance proposed in [24,34,46]. However, these approaches do not fully exploit the interaction between dropping and scheduling, and instead rely on admission control and traffic policing to enforce service differentiation.…”

IEEE Communications Surveys & Tutorials - third quarter 2008

“…This approach is adopted in [16,17], where the control does not necessarily make the allocated bandwidth reach a constant value as in [14,15]. We define the utilization as the fraction of time the server is busy.…”

“…In [17], we take a different approach by applying fuzzy control to maintain the average queue length between a queue threshold pair u th and l th that depends on the target loss rate. For a system that cannot be adequately described by detailed mathematical equations, fuzzy control is a convenient approach to synthesize a non-linear controller whose control laws are heuristically derived through some key insights of the process under control.…”

“…The closed-loop, or feedback control, approach arises naturally in this context where the packet loss, average queue length, or other system states, are regularly observed to provide the feedback to adjust the allocated bandwidth. The closed-loop control approach can be further categorized based on the guaranteed QoS metrics, including the average queue length [12,13], loss [14][15][16][17], and delay [18,19]. The open-loop control approach involves predicting the input traffic rate using the past history.…”

“…In real networks, however, the input traffic usually has complex (e.g.,ˆ.ˆˆ)ˆˆ( [12], [13] Queue length [23] Loss [14][15][16][17] Delay [18] Delay and rate [19] Guaranteed QoS metrics Zero loss [20][21][22] non-stationary, long-range dependent) and unknown characteristics, and thus the queuing process can have radically different characteristics from that of the M/M/1/K queue. As shown in Fig.…”