NETWORK-TRAFFIC PREDICTOR AND METHOD

• US 20190312802A1
• Filed: 04/04/2019
• Published: 10/10/2019
• Est. Priority Date: 04/04/2018
• Status: Active Grant

• Current Assignee: Cable Television Laboratories Incorporated
• Original Assignee: Cable Television Laboratories Incorporated
• Inventors: Huberman, Bernardo, Clearwater, Scott H.
• Granted Patent: US 10,778,555 B2
• Time in Patent Office: Days
• CPC Class Codes: H04L 41/0896 Bandwidth or capacity management, i.e. automatically increasing or decreasing capacities flow or congestion control using dynamic resource allocation, e.g. in-call renegotiation, H04L47/76
• CPC Class Codes: H04L 41/147 for predicting network behaviour
• CPC Class Codes: H04L 41/5025 by proactively reacting to service quality change, e.g. by reconfiguration after service quality degradation or upgrade
• CPC Class Codes: H04L 43/0882 Utilisation of link capacity
• CPC Class Codes: Y04S 40/00 Systems for electrical power generation, transmission, distribution or end-user application management characterised by the use of communication or information technologies, or communication or information technology specific aspects supporting them

Abstract

A method for predicting network-traffic bursts includes identifying, in data received by a networking device, a plurality of network-traffic bursts, each of the plurality of network-traffic bursts occurring at a respective one of plurality of burst-times {t_N, t_N-1, . . . , t₀}. The method includes determining a time-interval τ_n of a next burst occurring at τ_n after burst-time t₁ by determining respective values of τ_n, a parameter ξ, and a parameter η, that minimize, to within a tolerance, a quantity (f_k (ξ, η, k)−(τ_n−t_k)) for at least three values of a integer k. Parameters ξ and η are, respectively, a real and imaginary part of a power-law exponent of a power law relating predicted time-interval τ_n to any of the plurality of burst-times. The method includes determining, from a cumulative distribution function of a normal distribution of previously-identified network-traffic bursts, a time-duration during which the networking device may reallocate bandwidth.

Claims

1. A method for predicting network-traffic bursts comprising:

identifying, in data received by a networking device, a plurality of network-traffic bursts, each of the plurality of network-traffic bursts occurring at a respective one of plurality of burst-times {t_N, t_N-1, . . . , t₀}, N≥

3;

determining a time-interval τ

_n of a next burst occurring at τ

_n after burst-time t₁ by determining respective values of predicted time-interval τ

_n, a parameter ξ

, and a parameter η

that minimize, to within a tolerance, a quantity (f_k (ξ

, η

, k)−

(τ

_n−

t_k)) for three values of a positive integer k≤

N, parameters ξ and

η

being, respectively, a real part and an imaginary part of a power-law exponent of a power law relating predicted time-interval τ

_n to any of the plurality of burst-times; and

determining, from a cumulative distribution function of a normal distribution of previously-identified network-traffic bursts occurring within a time-interval that includes a previously-predicted burst-time τ

_p similar to predicted time-interval τ

_n by less than a predetermined tolerance, a time-duration during which the networking device may reallocate bandwidth according to at least one of traffic type, a subnet mask, and IP address.

2. The method of claim 1, f_k (ξ

, η

, k) being proportional to exp[η

<sup>−

1</sup> arctan(ξ

/η

)+kπ

/η

].

3. The method of claim 1, determining time-interval τ

_n including;

determining values ξ

_fit and η

_fit of parameters ξ and

η

as those that minimize, within a predetermined tolerance, a combination of |f_k (ξ

, η

, k)−

(t_n−

t_k)| for k=1, 2, . . . , N; and

determining τ

_n as an average of burst intervals τ

_k=t_k+f_k (ξ

_fit, η

_fit, k), for k=1, 2, . . . , N.

4. The method of claim 1, determining the time-duration comprising:

determining the time-duration from a mean burst-time Ip and a standard deviation σ

_p of the normal distribution.

5. The method of claim 4, further comprising determining the time-duration as √

{square root over (2)}σ

_p erf<sup>−

1</sup>(1−

2QoS)+μ

_p, where QoS is a predetermined metric related to a probability of having sufficient bandwidth during the time-duration.

6. The method of claim 4, further comprising determining the mean burst-time μ

_p and the standard deviation σ

_p by curve-fitting the previously-identified network-traffic bursts.

7. The method of claim 1, further comprising reallocating bandwidth during the time-duration and according to at least one of traffic type, a subnet mask, and IP address during the time-duration.

8. The method of claim 1, the network device being communicatively coupled to a processor and a memory storing non-transitory computer-readable instructions that, when executed by the processor, control the processor to perform the identifying, predicting, and determining.

9. The method of claim 1, the previously-predicted burst-time τ

_p being at least one of (i) different from predicted time-interval τ

_n by less than a predetermined tolerance and (ii) the one of a quantity of previously-predicted intervals τ

that is closest to predicted burst-interval τ

_n.

10. A network-traffic burst predictor comprising:

a processor communicatively coupled to a networking device; and

a memory storing non-transitory computer-readable instructions that, when executed by the processor, control the processor to;

identify, in data received by a networking device, a plurality of network-traffic bursts, each of the plurality of network-traffic bursts occurring at a respective one of plurality of burst-times {t_N, t_N-1, . . . , t₀}, N≥

3;

determine a time-interval τ

_n of a next burst occurring at τ

_n after burst-time t₁ by determining respective values of predicted time-interval τ

_n, a parameter ξ

, and a parameter η

that minimize, to within a tolerance, a quantity (f_k (ξ

, η

, k)−

(τ

_n−

t_k)) for three values of a positive integer k≤

N, parameters ξ and

η

being, respectively, a real part and an imaginary part of a power-law exponent of a power law relating predicted time-interval τ

_n to any of the plurality of burst-times; and

determine, from a cumulative distribution function of a normal distribution of previously-identified network-traffic bursts occurring within a time-interval that includes a previously-predicted burst-time τ

_p similar to predicted time-interval τ

_n by less than a predetermined tolerance, a time-duration during which the networking device may reallocate bandwidth according to at least one of traffic type, a subnet mask, and IP address.

11. The network-traffic burst predictor of claim 10, the networking device being at least one of a network node, a gateway, a router, a network switch, a network hub, and a base transceiver station.

12. The network-traffic burst predictor of claim 10, f_k (ξ

, η

, k) being proportional to exp[η

<sup>−

1</sup> arctan(ξ

/η

)+kπ

/η

].

13. The network-traffic burst predictor of claim 10, the memory further storing non-transitory computer-readable instructions that, when executed by the processor, control the processor to:

determine values ξ

_fit and η

_fit of parameters ξ and

η

as those that minimize, within a predetermined tolerance, a combination of |f_k (ξ

, η

, k)−

(t_n−

t_k)| for k=1, 2, . . . , N; and

determine τ

_n as an average of burst intervals τ

_k=t_k+f_k (ξ

_fit, η

_fit, k), for k=1, 2, . . . , N.

14. The network-traffic burst predictor of claim 10, the memory further storing non-transitory computer-readable instructions that, when executed by the processor, control the processor to:

determine the time-duration from a mean burst-time μ

_p and a standard deviation σ

_p of the normal distribution.

15. The network-traffic burst predictor of claim 14, the memory further storing non-transitory computer-readable instructions that, when executed by the processor, control the processor to:

determine the time-duration as √

{square root over (2)}σ

_p erf<sup>−

1</sup>(1−

2QoS)+μ

_p, where QoS is a predetermined metric related to a probability of having sufficient bandwidth during the time-duration.

16. The network-traffic burst predictor of claim 14, the memory further storing non-transitory computer-readable instructions that, when executed by the processor, control the processor to:

determine the mean burst-time μ

_p and the standard deviation σ

_p by curve-fitting the previously-identified network-traffic bursts.

17. The network-traffic burst predictor of claim 10, the memory further storing non-transitory computer-readable instructions that, when executed by the processor, control the processor to:

reallocate bandwidth during the time-duration and according to at least one of traffic type, a subnet mask, and IP address during the time-duration.

18. The network-traffic burst predictor of claim 10, the previously-predicted burst-time τ

_p being at least one of (i) different from predicted time-interval τ

_n by less than a predetermined tolerance and (ii) the one of a quantity of previously-predicted intervals τ

that is closest to predicted burst-interval τ

_n.