Processing

Please wait...

Settings

Settings

Goto Application

1. WO2020142407 - SYSTEM AND METHOD FOR ESTIMATION OF QUALITY OF EXPERIENCE (QOE) FOR WEB BROWSING USING PASSIVE MEASUREMENTS

Note: Text based on automatic Optical Character Recognition processes. Please use the PDF version for legal matters

[ EN ]

WHAT IS CLAIMED IS:

1. A computer-implemented method for predicting quality of experience (QoE) performance of an application downloading a webpage over a network, comprising:

determining a stat data corresponding to a traffic through a network node of a network path between the application and a server stack;

generating a feature vector, based at least in part on at least a portion of the stat data; estimating a download performance metric for the application downloading the webpage from one or more servers in the server stack, based at least in part on at least a portion of the feature vector; and

estimating a QoE value, based at least in part on the estimated performance metric.

2. The computer-implemented method of claim 1 for predicting QoE performance of an application downloading a webpage over a network, wherein

estimating the download performance metric is based, at least in part, on inputting the feature vector to a machine learning (ML) model, a statistical model, or both.

3. The computer-implemented method of claim 1 for predicting QoE performance of an application downloading a webpage over a network, wherein:

the stat data includes

connection-level stat data indicating a plurality of different connections, each being between the application and a respective server for the connection, the server being within the server stack, and

application-level stat data, indicating transfers of webpage content, over respective ones of the different connections, each transfer being from the respective server for the connection to the application; and

generating the connection-level stat data is based, at least in part, on measurements at the network node.

4. The computer-implemented method of claim 3 for predicting QoE performance of an application downloading a webpage over a network, wherein:

the application-level stat data includes object-level stat data, and

generating the object-level stat data includes estimating at least a portion of the object-level stat data, based at least in part on least a portion of the connection-level stat data.

5. The computer-implemented method of claim 3 for predicting QoE performance of an application downloading a webpage over a network, wherein:

the connection-level stat data indicates, for each of at least a sub-plurality of the connections, a connection start time, a network address of the webpage access application, a network address of the respective server for the connection, and a quantity of webpage content packets that have been communicated over the connection, from the server for the connection to the webpage access application.

6. The computer-implemented method of claim 3 for predicting QoE performance of an application downloading a webpage over a network, wherein:

determining the stat data corresponding to the traffic through the network node includes detection of Transport Layer Security (TLS) record layer information,

the application-level stat data includes object-level stat data, and

generating the object-level stat data is based at least in part on the detection of TLS record layer information.

7. The computer-implemented method of claim 3 for predicting QoE performance of an application downloading a webpage over a network, wherein generating the feature vector includes:

generating a sequence of feature data, by operations that include, over each time interval in a sequence of time intervals, determining a new connection-level stat data or updated connection-level stat data in response to said measurements indicating a respective end of one or more of the connections, a respective start of one or more new ones of the connections, or both, and

generating the feature vector, based at least in part on a concatenating of integer N feature data among the sequence of feature data.

8. The computer-implemented method of claim 7 for predicting QoE performance of an application downloading a webpage over a network, wherein

estimating the download performance metric is based, at least in part, on inputting the feature vector to a machine learning (ML) model, inputting the feature vector to a statistical model, or both, and

outputting the estimated download performance metric based, at least in part, on a result of said inputting the feature vector.

9. The computer-implemented method of claim 7 for predicting QoE performance of an application downloading a webpage over a network, wherein the estimation model is configured to:

estimate whether the feature vector corresponds to page load activity during the time window or to no page load activity during the time window and, based at least in part on a result, output the estimated download performance metric based.

10. The computer-implemented method of claim 7 for predicting QoE performance of an application downloading a webpage over a network, wherein a concatenating of the N time intervals corresponding, respectively, to the N feature data integer forms a time window, and wherein estimating the performance metric includes:

inputting the feature vector to an estimation model,

generating a download activity estimate, based at least in part on a result of an inputting of the feature vector to the estimation model, wherein the download activity estimate indicates whether download activity is estimated to occur or to not occur during the time window, and

generating a value of the estimated download performance metric, based at least in part on the download activity estimate.

11. The computer-implemented method of claim 10 for predicting QoE performance of an application downloading a webpage over a network, wherein estimating the performance metric includes:

generating an updated feature vector, by operations that include sliding the time window at least one of the intervals, and basing the update at least in part on new feature data corresponding to the at least one of the time intervals;

inputting the updated feature vector to the estimation model;

generating a new download activity estimate, based at least in part on a result of the inputting of the updated feature vector to the estimation model, wherein:

the new download activity estimate indicates whether download activity is estimated to occur or to not occur during the slid time window, and

generating the value of the estimated download performance metric is further based, at least in part, on the new download activity estimate.

12. The computer-implemented method of claim 11 for predicting QoE performance of an application downloading a webpage over a network, wherein estimating the performance metric includes:

generating a sequence of new download activity estimates, by operations that include repeating, for an integer number of repetitions:

generating another updated feature vector, by operations that include sliding the time window another at least one of the intervals, and basing the update at least in part on other new feature data corresponding to the another at least one of the time intervals,

inputting the another updated feature vector to the estimation model, and

generating another new download activity estimate, based at least in part on a result of the inputting of the another updated feature vector to the estimation model, and the another new download activity estimate indicates whether download activity is estimated to occur or to not occur during the another slid time window,

wherein generating the value of the estimated download performance metric is further based, at least in part, on the another new download activity estimate.

13. The computer-implemented method of claim 11 for predicting QoE performance of an application downloading a webpage over a network, wherein estimating the performance metric includes:

generating a sequence of new raw download activity estimates, by operations that include repeating, for an integer number of repetitions:

generating another updated feature vector, by operations that include sliding the time window another at least one of the intervals, and basing the update at least in part on other new feature data corresponding to the another at least one of the time intervals,

inputting the another updated feature vector to the estimation model, and generating another new raw download activity estimate, based at least in part on a result of the inputting of the another updated feature vector to the estimation model, and the another new raw download activity estimate indicates whether download activity is estimated to occur or to not occur during the another slid time window;

applying an error correction process to at least a portion of the sequence of new raw download activity estimates; and

based at least in part on a result of applying the error correction process, generating a sequence of new download activity estimate,

wherein generating the value of the estimated download performance metric is further based, at least in part, on at least a portion of the sequence of new download activity estimates.

14. A system, comprising:

a processor, coupled to a node of an access path between a wide area network (WAN) and an external interface configured to interface with a webpage access device;

a memory, coupled to the processor, storing processor-executable instructions that, when executed by the processor, cause the processor to:

detect transport-layer connections extending through the node, the connections having at least a likelihood of association with a downloading, by a webpage access application associated with the webpage access device, of a webpage hosted by a server stack and, in response, generate connection-level stat data, the connection-level stat data indicating, for each of the connections, a connection start time, a connection end time, a network address for a respective server in the server stack, and a network address for the webpage access application,

generate a feature vector, based at least in part on at least a portion of the connection-level stat data,

estimate a download performance metric for the webpage access application downloading the webpage from the server stack, based at least in part on at least a portion of the feature vector; and

estimate a QoE value, based at least in part on the estimated performance metric.

15. The system of claim 14, wherein the storing processor-executable instructions include processor-executable instructions that, when executed by the processor, cause the processor to:

estimate application-level stat data that indicates a plurality of transfers, or likely transfers, or both, of a plurality of different contents of the webpage, over the transport-layer connections,

wherein:

the estimated application-level stat data includes, for each estimated transfer or likely transfer, an identification of the respective server, a time stamp, and an identifier of the webpage content, and

generating the feature vector is further based, at least in part, on at least a portion of the estimated application-level stat data.

16. The system of claim 14, wherein the processor-executable instructions include processor-executable instructions that, when executed by the processor, cause the processor to:

implement an estimation model, the estimation model including a machine learning (ML) model, a statistical model, or both, and

estimate the download performance metric, further based at least in part on inputting the at least a portion of the feature vector to the estimation model.

17. The system of claim 14, wherein:

the application-level stat data includes object-level stat data that indicates the different contents of the webpage as different webpage objects and indicates, for each of the different webpage objects, an object identifier,

the time stamp is a request time stamp, indicating a time of a request sent by the webpage access application, for the object, and

the object-level stat date further includes, for each of the different webpage objects, a time of the respective server starting a response to the request.

18. The system of claim 14, wherein the processor-executable instructions include processor-executable instructions that, when executed by the processor, cause the processor to:

detect Transport Layer Security (TLS) record layer information, and

estimate the application-level stat data based, at least in part, on the detection of TLS record layer information.

19. A computer-implemented method for predicting quality of experience (QoE) performance of a webpage access application downloading a webpage over a network, comprising:

detecting transport-layer connections extending through a node, the connections having at least a likelihood of association with a downloading, by the webpage access application of the webpage, a server stack and, in response, generating connection-level stat data that indicates, for each of the connections, a connection start time, a connection end time, a network address for a respective server in the server stack, and a network address for the webpage access application;

generating a feature vector, based at least in part on at least a portion of the connection-level stat data;

estimating a download performance metric for the webpage access application downloading the webpage from the server stack, based at least in part on at least a portion of the feature vector; and

estimating a QoE value, based at least in part on the estimated performance metric.