Resilient Buffer Allocation Scheme in Active Queue Management: Stochastic Cooperative Game Theoretic Approach

During the last decade, a plentiful number of active queue management (AQM) schemes have been proposed, but their main objectives are to simply allocate the buffer resource to all flows evenly, or prioritize responsive flows over irresponsive flows. We note that the sending rates of responsive flows can be determined diversely, and not all irresponsive flows result in high sending rates. Furthermore, it is rational to exclude the buffer resource allocated to certain privileged traffics. Grounded by these evidences, in this paper, we present a resilient AQM algorithm, named Prior-Core-based Buffer Allocation (PCBA) considering the diverse congestion control algorithms, the requirements of all flows regardless of the responsiveness, and the existence of privileged traffics. Our approach is based on stochastic cooperative game theory, where the payoffs yielded by the cooperation are described by random variables, and the core can be defined only over the distribution of these random payoffs; the core in this situation is called prior-core. Our main objectives are to define the attitude toward risk of each coalition along with the context of fair buffer allocation, and then to achieve the buffer allocation placed in the prior-core. It is shown that the buffer allocation, yielded by the prior-core and under the defined attitude toward risk, realizes that the smaller the buffer requirement is, the more preferentially it is served regardless of the responsiveness, and the buffer allocation for aggressive flows would be restricted; all these are verified through the simulation experiments.