US 2020/0092622 A1 · METHOD AND APPARATUS FOR ALLOCATING BANDWIDTH BASED ON MACHINE LEARNING IN PASSIVE OPTICAL NETWORK

Abstract

A method and apparatus for allocating a bandwidth based on machine learning in a passive optical network, the method including generating an inference model to predict a consumed bandwidth required for transmission by learning unstructured data of a PON including an OLT and traffic data corresponding to state information of the PON collected from the PON, predicting a consumed bandwidth with respect to a queue corresponding to a class requiring a low-latency service among classes of ONUS connected to the OLT based on the generated inference model, performing a VBA with respect to the queue corresponding to the class requiring the low-latency service based on the predicted consumed bandwidth, and performing a DBA with respect to a queue corresponding to a class not requiring the low-latency service using a transmission bandwidth which remains after the VBA is performed.

Claims18

Dependent claims

A bandwidth allocating method performed by a bandwidth allocating apparatus included in an optical line terminal (OLT), the bandwidth allocating method comprising:

generating an inference model to predict a consumed bandwidth required for transmission by learning, using a plurality of machine learning models, unstructured data of a passive optical network (PON) including the OLT and traffic data corresponding to state information of the PON collected from the PON;
predicting a consumed bandwidth with respect to a queue corresponding to a class requiring a low-latency service among classes of optical network units (ONUs) connected to the OLT based on the generated inference model;
performing a variable bandwidth allocation (VBA) with respect to the queue corresponding to the class requiring the low-latency service among classes of the ONUs based on the predicted consumed bandwidth;
and
performing a dynamic bandwidth allocation (DBA) with respect to a queue corresponding to a class not requiring the low-latency service among the classes of the ONUs using a transmission bandwidth which remains after the VBA is performed.

2 more elements

A bandwidth allocating apparatus included in an optical line terminal (OLT), the bandwidth allocating apparatus comprising:

a machine learning inferrer configured to generate an inference model to predict a consumed bandwidth required for transmission by learning, using a plurality of machine learning models, unstructured data of a passive optical network (PON) including the OLT and traffic data corresponding to state information of the PON collected from the PON, and predict a consumed bandwidth with respect to a queue corresponding to a class requiring a low-latency service among classes of optical network units (ONUs) connected to the OLT based on the generated inference model;
and
a bandwidth allocator configured to perform a variable bandwidth allocation (VBA) with respect to a queue corresponding to a class requiring the low-latency service among classes of the ONUS based on the predicted consumed bandwidth, and perform a dynamic bandwidth allocation (DBA) with respect to a queue corresponding to a class not requiring the low-latency service among the classes of the ONUs using a transmission bandwidth which remains after the VBA is performed.

A bandwidth allocating method performed by a bandwidth allocating apparatus of an optical network unit (ONU) connected to an optical line terminal (OLT), the bandwidth allocating method comprising:

storing uplink data for each of different queues for services having priorities;
transmitting a report message requesting a bandwidth allocation to the OLT based on the stored uplink data;
receiving a grant message including a transmission bandwidth allocated based on the transmitted report message;
and
transmitting the stored uplink data based on the received grant message,
wherein the grant message includes a transmission bandwidth allocated by performing a variable bandwidth allocation (VBA) with respect to queues corresponding to a class requiring a low-latency service within a bandwidth allocable according to a cycle time, and performing a dynamic bandwidth allocation (DBA) with respect to queues corresponding to a class not requiring the low-latency service.

3 more elements

Dependencies are inferred from the claim text.

Method and apparatus for allocating bandwidth based on machine learning in passive optical network

Life of this patent

Method and apparatus for allocating bandwidth based on machine learning in passive optical network

Life of this patent

Abstract

Claims18