Active Queue Management Algorithm For Time Demand

With the popularization and development of the network intelligent terminal,higher requirements and challenges have been raised about congestion control of the network and quality of service based on delay and jitter.The queue management of intermediate network nodes plays a vital role in improving network congestion.Because of deadlock and global synchronization,passive queue management is replaced by active queue management which uses various network measures and loses packets in advance.Existing active queue management methods generally use network measurement to get queue length and queue delay,and discard some data packets in advance according to the set threshold to avoid network congestion.However,the existing active queue management methods do not distinguish the importance of the packet itself when discarding the data packet,and neglect the QoS requirement of different application data packets for delay.Based on the analysis and research of active queue management for time delay control,an active queue management method for time demand is proposed to improve the above shortcomings.The basis of queue management is adjusted to the datagram itself,and the time demand is used to control the entry and exit of data packets.This paper mainly includes the following contents:(1)Aiming at the problem that existing methods neglect the requirement of data packet itself and discard them uniformly,a parameter named "time demand" is defined,which is the longest time that data packets exist in routers.The entry strategy TD-AQM for time demand is proposed.TD-AQM calculates the estimated queue location of data packets through time demand,and constructs a two-dimensional queue using "occupancy","virtual occupancy" and "side hang" flags.Each data packet is queued or discarded according to the queue entry strategy.Compared with CoDel and other mainstream methods,it was proved that TD-AQM performed better in throughput rate and other measures,and met the packet time demands.(2)According to the experimental results of TD-AQM,the phenomenon of "flood effect" in queue management is summarized.The flood effect is quantitatively analyzed through experiments,and RFTDA,an active queue management method for time demand,is proposed to restrain the flood effect.By limiting the downward search range and probability to contain the virtual occupancy hit rate of the cell,the method ensures that the data packets with urgent time demand have sufficient occupancy resources.The experimental results showed that RFTDA can effectively restrain the flood effect when it meets the time demand of incoming packets.(3)Based on the above methods,the performance verification system of RFTDA is implemented.The system measures various measures of RFTDA,such as throughput rate and delay satisfaction,by constructing multiple flow with different probability distribution delay requirements.The results showed that RFTDA performed well in different experimental groups.Using the interface display module,the information status of queues and cells can be visualized in real time through the interaction of kernel state and user state,which shows that the system has strong practicability.