Measurement-Based Admission Control: A Revisit

Measurement-based admission control (MBAC) is an important technique for providing statistical service guarantees and many MBAC algorithms have been proposed in the literature. It has also been investigated in a previous work that these algorithms achieve nearly the same utilization for a given packet loss rate or in short the same loss-load curve. Based on this investigation, it has been argued that further research on better MBAC equations will likely be a fruitless endeavor since there is little room to improve existing algorithms in terms of utilization. In this paper, we investigate why the MBAC algorithms investigated in the previous work have the same loss-load curve. We find this not surprising. We hence argue that it should be other issues than the loss-load curve that should be taken into account in designing an MBAC algorithm. Based on this argument, we revisit the assumptions made by these MBAC algorithms and examine the rationales of these assumptions when these MBAC algorithms are used in different network scenarios. Finally, we discuss some remaining challenges for MBAC research.