A Demand Partitioning Based Topology Aggregation Mechanism for Scalable QoS Guarantee

In addressing the scalability and accuracy tradeoff in topology aggregation mechanisms with delay and bandwidth constraints in hierarchical networks, we propose a demand partitioning based topology aggregation mechanism called dominative points based topology aggregation (DPTA). The main drawback of traditional approaches is that these approaches use lines or curves to approximately represent the staircase which consists of representative points so that the performance degrades obviously as the number of representative points increases. However, DPTA uses dominative points to construct a new staircase according to the demand partitioning, so as to obtain the O(1) space complexity and avoid performance degradation as the network expands. Firstly, we perform demand partitioning for the arrival requests. Secondly, the representative points are further aggregated to fewer amounts of dominative points. Based on the dominative points, DTPA eliminates the influence of the inaccuracy information accumulation on the performance of the topology aggregation process. Lastly, optimal allocation of available bandwidth is presented to maximize the number of accepted requests. Simulation results show that the proposed mechanism can guarantee lower delay deviation, higher success ratio with lower crankback ratio and accept more requests than the traditional approaches.