In CSFQ, when there exist adaptive flows there will be error between the measurement value of throughput and its real value, furthermore the fairness is seriously degraded. To resolve the problem, the paper mainly studies that how to improve the accuracy of the computation of a value, and proposes a queue length embedded core-stateless fair queue algorithm (QL-CSFQ). In QL-CSFQ, the queue length is concerned to improve the accuracy of the computation of value. The paper compares QL-CSFQ to CSFQ in fairness, utilization of bandwidth, the queue length and its jitter using large amount of simulations, and the results show that QL-CSFQ can decrease the impact of measurement error on the fairness.
