System for real-time prediction of quality for internet-based multimedia communications

• US 6,665,271 B1
• Filed: 03/17/1999
• Issued: 12/16/2003
• Est. Priority Date: 03/17/1998
• Status: Expired due to Term

• Current Assignee: Transnexus Incorporated
• Original Assignee: Transnexus Incorporated
• Inventors: Martinez, John V., Dalton, James P. G. Jr., de Figueiredo, Rui J. P., Decker, David B., Thomas, Stephen A.
• Primary Examiner(s): Vu, Huy D.
• Assistant Examiner(s): RYMAN, DANIEL J
• Application Number:
  US09/270,978
• Time in Patent Office: 1,735 Days
• Field of Search: 370/252, 370/401, 370/352, 370/260, 370/231, 370/351, 370/237, 370/400, 370/470, 379/111, 379/115, 379/114, 379/220, 379/130, 379/112, 379/242, 709/104, 709/223
• US Class Current: 370/252
• CPC Class Codes: H04L 1/20 using signal quality detector

Abstract

Predicting the quality of a communication carried via a distributed computer network, such as the global Internet, before the initiation of that communication between a pair of endpoints. A prediction system comprises software-implemented processes that can be implemented within an Internet gateway or in an external system that provides prediction results to an Internet gateway. A first process accepts historical data and current environment data and, in response, generates a prediction of objective performance characteristics. A second process accepts the objective performance characteristics and, in response, generates an estimate of the expected quality of communications with an end user. This user quality estimate can be used to support a determination of whether to proceed with the communication or to rely upon an alternative to computer-network supported communications.

Claims

1. A system for generating a prediction result for quality of a multimedia communication for a distributed computer network comprising:

an objective performance prediction module, responsive to historical communication data and current communication environment data, said historical and current data being passively measured, for generating a prediction of objective performance characteristics, wherein the objective prediction module comprises;

a genetic program module, responsive to the historical communication data and the current communication environment data, for performing a feature selection by selecting relevant portions of the historical communication data, wherein the genetic program module is tuned to select the relevant portions of the historical communication data based on selection criteria comprising similarity of network communication paths, the module being also tuned to identify endpoints for the communication and time and date information for the communication; and

a neural network, responsive to the current communication environment data and the relevant portions of the historical communication data, for generating the prediction of objective performance characteristics for the communication, wherein the neural network comprises an optimal interpolative neural network that generates the prediction of the objective performance characteristics, the neural network comprising;

an input layer comprising inputs β

(n−

1);

an epoch layer comprising neurons, each having a synaptic weight represented by the vector component of a vector j(n−

1) and responsive to the inputs β

(n−

1) for generating a neuron output; and

an output variable layer comprising a linear neuron for generating a predicted value of u(n)=M⁴(n) in response to the output of each neuron on the epoch layer; and

an estimation module, responsive to the prediction of objective performance characteristics, for generating an estimate of expected quality of the communication.

2. The system of claim 1, wherein the objective performance prediction module is operative to collect the historical communication data and the current communication environment data by using voice over internet protocol network signaling protocols, call detail records and usage indication messages to obtain prior communication measurements and attributes of the current communication environment.

3. The system of claim 1, wherein the objective prediction module is operative to collect the historical communication data and the current communication environment data by acquiring fine-grained path information and coarse-grained path information as indicators of the quality of service for the communication between a pair of endpoints.

4. The system of claim 1, wherein the estimation module comprises a neural network, responsive to the prediction of objective performance characteristics, for generating the estimate of expected quality of the communication.

5. The system of claim 1, wherein the estimation module generates the estimate of expected quality of the communication in response to the prediction of objective characteristics comprising a fraction packet loss for the communication, as measured from each of the endpoints for the communication, and a round-trip delay for the communication, as measured from each of the endpoints.

6. The system of claim 1, wherein the estimation module comprises a model constructed such that objective performance is correlated with an expected quality of service perceived by the end user.

7. The system of claim 1, wherein the historical communication data comprises:

attributes of the communication, comprising identities of the endpoints for the communication and time and day for the communication;

communication-related statistics comprising total packet loss and fractional packet loss for each direction of the communication;

characterizations of both the round-trip delay between the endpoints for the communication and the fine-grained network path from the endpoint responsible for initiating the communication; and

characterization of the coarse-grained network path from each of the endpoints for the communication.

8. The system of claim 1, wherein the current communication environment data comprises:

attributes of the communication, comprising identities of the endpoints for the communication and time and day for the communication;

characterizations of the round-trip delay from the endpoint responsible for initiating the communication; and

characterization of the coarse-grained network path from each of the endpoints for the communication.

9. A computer-implemented process for generating a prediction result for quality of a communication carried over a distributed computer network, comprising the steps:

in response to historical communication data, selecting relevant portions of the historical data based on selection criteria comprising similarity of network communication paths, identity of endpoints for the communication, and time and date information for the communication using a genetic program module;

in response to the relevant portions of historical communication data and current communication environment data, passively generating a prediction of objective performance characteristics using an optimal interpolative neural network, the neural network comprising;

an input layer comprising inputs β

(n−

1);

an epoch layer comprising neurons, each having a synaptic weight represented by the vector component of a vector j(n−

1) and responsive to the inputs β

(n−

1) for generating a neuron output; and

an output variable layer comprising a linear neuron for generating a predicted value of u(n)=M⁴(n) in response to the output of each neuron on the epoch layer; and

in response to the prediction of objective performance characteristics, generating an estimate of expected quality of the communication.

10. The computer-implemented process of claim 9, further comprising the step of obtaining the historical communication data by collecting a measurement of a round-trip time between communicating endpoints from communication call detail records and usage indication messages.

11. The computer-implemented process of claim 10, wherein the round trip time is defined by a time period extending between a transmission by one of the endpoints of an IRQ message and a reception of a corresponding IRR response by the endpoint.

12. The computer-implemented process of claim 9 further comprising the step of obtaining the historical communication data by using network peering data to collect coarse-grained network path information for the communication.

13. The computer-implemented method of claim 9, wherein the historical communication data comprises;

attributes of the communication, comprising identities of the endpoints for the communication and time and day for the communication;

communication-related statistics comprising total packet loss and fractional packet loss for each direction of the communication;

characterizations of both the round-trip delay between the endpoints for the communication and the fine-grained network path from the endpoint responsible for initiating the communication; and

characterization of the coarse-grained network path from each of the endpoints for the communication.

14. The computer-implemented method of claim 9, wherein the current communication environment data comprises:

attributes of the communication, comprising identities of the endpoints for the communication and time and day for the communication;

characterizations of the round-trip delay from the endpoint responsible for initiating the communication; and

characterization of the coarse-grained network path from each of the endpoints for the communication.

15. The computer-implemented method of claim 9 further comprising the step of determining whether to proceed with the communication over the distributed network based on the estimate of expected quality of the communication.