RNN-based EPON dynamic bandwidth allocation algorithm

Abstract

The invention relates to an EPON dynamic bandwidth allocation algorithm based on RNN. The method comprises the following steps: firstly, an ONU reports a bandwidth request to an OLT; then the OLT judges whether the service is an EF service or not, carries out bandwidth prediction and bandwidth allocation on the EF service, generates actually required bandwidths for the bandwidth requests of the AF service and the BE service, and carries out bandwidth allocation; then the remaining bandwidth requests of the AF and BE services are summarized to carry out bandwidth allocation in a queuing scheduling mode. In the invention, the bandwidth prediction and bandwidth allocation are carried out on the high-priority service EF on the basis of the RNN structure model, prediction accuracy is improved, bandwidth allocation is carried out on the AF service and the BE service on the premise that the high-priority service is ensured, and finally bandwidth allocation is carried out on the remaining bandwidth requests of the three services in a queuing scheduling mode; therefore, the service quality of other services is ensured to a certain extent, the prediction accuracy is improved, the queue delay is reduced, and the uplink bandwidth utilization rate is improved.

Description

technical field [0001] The invention belongs to the technical field of bandwidth utilization, and in particular relates to an RNN-based EPON dynamic bandwidth allocation algorithm. Background technique [0002] RNN is a kind of recurrent neural network with memory ability, which can obtain prediction ability through prior knowledge learning, and through online learning, modify parameters in real time to improve the prediction accuracy of nonlinear data. The process of sending data from ONU to OLT is as follows: figure 1 As shown, during the waiting period of T1-T5, the ONU will still receive uplink data, and this part of the data volume is not included in the request frame, causing the phenomenon that the request information does not match the actual demand, so this part of the data packets can only be accumulated Sending until the next polling cycle not only increases the average queue delay, but also affects the real-time performance of the business. [0003] Compared wi...

AI Technical Summary

Problems solved by technology

The process of sending data from ONU to OLT is as follows: figure 1 As shown, during the waiting period of T1-T5, the ONU will still receive uplink data, and this part of the data volume is not included in the request frame, causing the phenomenon that the request information does not match the actual demand, so this part of the data packets can only be accumulated Sending until the next polling cycle not only increases the average queue delay, but also affects the real-time performance of the business

Images