Using machine learning techniques to predict device buffer status
Devices, systems, methods, and processes are described herein that can reduce network communication overhead by reducing data that is sent. This reduction in data can be achieved by reducing the requests for data related to the buffers of one or more network devices. This buffer status data reduction can be achieved by predicting the current status of the buffers in other network devices. When the buffer predictions are sufficient, the need to request or poll for buffer status reports can be reduced. The buffer predictions can be evaluated through one or more confidence levels. These buffer predictions and confidence levels can be generated by one or more machine learning processes that are configured to operate on one or more network devices. Based on the confidence levels of the predictions, the number of buffer status report polls and of any unsolicited buffer service reports can be reduced.
1 . A device, comprising:
a processor;
at least one network interface controller configured to provide access to a plurality of devices over a network; and
a memory communicatively coupled to the processor, wherein the memory comprises a buffer prediction logic configured to:
couple to a plurality of network devices, wherein each of the plurality of network devices comprises at least one buffer;
issuing a plurality of buffer service report (BSR) polls at a first polling rate;
receiving at least one BSR, wherein the BSR comprises BSR data;
utilize one or more machine learning processes to generate a confidence level associated with the buffer of at least one of the plurality of network devices;
update the first polling rate to a second polling rate based on the generated confidence level; and
receive ground truth data.
2 . The device of claim 1 , wherein the BSR polls are issued directly to the plurality of network devices.
3 . The device of claim 1 , wherein the one or more machine learning processes utilize BSR data to generate the confidence level.
4 . The device of claim 3 , wherein the BSR data is associated with a period of time.
5 . The device of claim 1 , wherein the one or more machine learning processes generates a prediction of a current buffer status.
6 . The device of claim 1 , wherein the ground truth data is incorporated into the BSR.
7 . The device of claim 1 , wherein the one or more machine learning processes are further configured to generate an updated confidence level based on at least a comparison between the ground truth data to a previously generated prediction.
8 . The device of claim 7 , wherein the updated confidence level is transmitted to at least one of the plurality of network devices.
9 . The device of claim 7 , wherein the buffer prediction logic is further configured to monitor the updated confidence level over a period of time.
10 . The device of claim 9 , wherein, in response to the updated confidence level falling below a predetermined threshold, the buffer prediction logic is further configured to increase the BSR polling rate.