Network performance monitor

• US 8,719,398 B2
• Filed: 02/29/2008
• Issued: 05/06/2014
• Est. Priority Date: 02/29/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Qian, Tin, Ramarathinam, Aravind, Bishop, Michael
• Primary Examiner(s): Desrosiers, Evans
• Application Number:
  US12/074,146
• Publication Number:
  US 20090222555A1
• Time in Patent Office: 2,258 Days
• Field of Search: 709223-224
• US Class Current: 709/224
• 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 41/5003 Managing SLA; Interaction between SLA and QoS
• CPC Class Codes: H04L 43/0864 Round trip delays
• CPC Class Codes: H04L 43/0882 Utilisation of link capacity

Abstract

A network performance monitor for a computer. The performance monitor maintains a baseline for communication paths between a networked computer and other networked computers connected. The network performance baseline is established and maintained by measuring and updating network performance parameters of the communication paths. The network performance parameters are updated using values obtained during time intervals when the communication path consumed a bandwidth approximately equal to its estimated bandwidth. If congestion is detected on the network or the network is underutilized, values of the network performance parameters obtained during that interval are not used to update the network performance parameters. Some performance parameters in the baseline, such as minimum and maximum bandwidth values may be updated by computing a moving average of each parameter. Asymmetric weight values for updating the moving average may be selected based on a direction of change of a value of a network parameter.

Claims

1. A method of monitoring network performance by maintaining at least one network performance parameter for at least one communications path between a first networked computer and at least one other networked computer connected via a network, the method comprising:

determining that, during a first time interval, a first bandwidth within a bandwidth range was utilized for communications on the at least one communications path;

obtaining at least one delay parameter for the at least one communications path over the first time interval, the at least one delay parameter being different than an estimated bandwidth;

for the first time interval, updating at least one stored value of the at least one network performance parameter in response to the determination that the first bandwidth was utilized for the communications on the at least one communications path during the first time interval, wherein the updating includes;

changing the at least one stored value based on the at least one delay parameter for the at least one communications path over the first time interval and based on at least one asymmetric weight, wherein the at least one asymmetric weight is based on relative values of the at least one delay parameter for the at least one communications path over the first time interval and at least one delay parameter for the at least one communications path over a previous time interval;

assessing the network performance based on the updated at least one stored value;

determining that, during a second time interval, a second bandwidth outside of the bandwidth range was utilized for communications on the at least one communications path; and

for the second time interval, retaining the at least one stored value, without change, in response to the determination that the second bandwidth was utilized for communications on the at least one communications path during the second time interval.

2. The method of claim 1, wherein determining that the first bandwidth was utilized for communications on the at least one communications path during the first interval comprises:

determining that the first bandwidth exceeds a threshold percentage of the estimated bandwidth.

3. The method of claim 2, wherein determining that the first bandwidth exceeds the threshold percentage of the estimated bandwidth comprises:

determining that a window set by a network stack had a size exceeding a threshold percentage of a value representing a product of the estimated bandwidth and an estimated round trip delay on the at least one communications path.

4. The method of claim 2, wherein determining that the first bandwidth was utilized for communications on the at least one communications path further comprises:

determining whether the at least one communications path is congested.

5. The method of claim 4, wherein determining whether the at least one communications path is congested comprises:

determining whether, during an additional time interval that includes the first time interval, variation in Round Trip Time (RTT) on the at least one communications path exceeded a RTT threshold.

6. The method of claim 1, further comprising:

identifying a network performance degradation based on a difference between the at least one delay parameter and the at least one stored value; and

in response to the identification of the network performance degradation, identifying at least one source of the network performance degradation.

7. The method of claim 6, wherein the network comprises a local area network, and wherein identifying the at least one source of the network performance degradation comprises:

obtaining network communication information from at least one second other networked computer connected to the local area network; and

determining that the at least one source of the network performance degradation comprises at least one selected from the group consisting of cross-traffic and a layer 2 dynamic problem.

8. The method of claim 6, wherein the network comprises a wide area network, and wherein identifying the at least one source of the network performance degradation comprises:

obtaining network communication information from at least one second other networked computer connected to the wide area network; and

determining that the at least one source of the network performance degradation comprises at least one selected from the group consisting of a remote server and an internet service provider.

9. The method of claim 1, wherein the at least one delay parameter comprises a bandwidth parameter computed as a window size for the at least one communications path divided by a round trip time for the at least one communications path.

10. A method of maintaining a network performance baseline for at least one communications path between a first networked computer and least one other networked computer connected via a network, the method comprising:

maintaining the network performance baseline by maintaining a moving average of at least one network performance parameter over a plurality of intervals, the at least one network performance parameter representing an extreme value of a bandwidth parameter, the maintaining comprising, for each of the plurality of intervals;

obtaining, from a plurality of measurements of the at least one network performance parameter, an extreme value during that interval; and

updating the moving average of the at least one network performance parameter, the updating comprising computing a combination of the extreme value for that interval weighted by a first factor and a value of the moving average computed for a prior interval weighted by a second factor, at least one of the first factor and the second factor being asymmetrically selected based on the extreme value for that interval relative to the moving average computed for the prior interval.

11. The method of claim 10, wherein:

the extreme value comprises a maximum bandwidth;

the extreme value during that interval comprises a maximum bandwidth during that interval; and

the moving average computed for the prior interval comprises a moving average of maximum bandwidth over the prior interval and one or more intervals preceding the prior interval.

12. The method of claim 11, wherein:

in response to the maximum bandwidth for that interval being higher than the moving average of the maximum bandwidth for the prior interval, selecting the first and second factors based on a first ratio;

in response to the maximum bandwidth for that interval not being higher than the moving average of the maximum bandwidth for the prior interval, selecting the first and second factors based on a second ratio; and

the first ratio is greater than the second ratio.

13. The method of claim 10, wherein:

the extreme value comprises a minimum bandwidth;

the extreme value during that interval comprises a minimum bandwidth during that interval; and

the moving average computed for the prior interval comprises a moving average of the minimum bandwidth over the prior interval and one or more intervals preceding the prior interval.

14. The method of claim 13, wherein:

in response to the minimum bandwidth for that interval being lower than the moving average of the minimum bandwidth for the prior interval, selecting the first and second factors based on a first ratio;

in response to the minimum bandwidth for that interval not being lower than the moving average of the minimum bandwidth for the prior interval, selecting the first and second factors based on a second ratio; and

the first ratio is greater than the second ratio.

15. The method of claim 10, further comprising maintaining a moving average of a mean bandwidth during the plurality of intervals, the maintaining the moving average of the mean average comprising, for each of the plurality of intervals:

obtaining, from a plurality of measurements of bandwidth, the mean bandwidth over that interval; and

updating the moving average of the mean bandwidth, the updating comprising computing a combination of the mean bandwidth during that interval weighted by a third factor and a value of the moving average of the mean bandwidth of the prior interval weighted by a fourth factor.

16. The method of claim 10, further comprising obtaining, for each of the plurality of intervals, path usage parameters indicating usage of the at least one communications path, the path usage parameters comprising at least one or more of a round-trip-time, a variance of the round-trip-time, a duration of usage of the at least one communications path, an amount of usage of the at least one communications path, and a duration of time elapsed since last usage of the at least one communications path.

17. A networked computer having a processor and a memory having stored thereon computer-executable modules for, in response to execution by the processor, maintaining a network performance parameter for at least one communications path between the networked computer and least one other networked computer connected via a network, the computer-executable modules comprising:

a network stack that tracks and provides, upon a request from an assessment module, at least one delay parameter for the at least one communications path for each of a plurality of time intervals;

an interface between the network stack and the assessment module that receives the at least one delay parameter from the network stack and provides the at least one delay parameter to the assessment module; and

the assessment module that;

determines whether, during each of the plurality of time intervals, bandwidth within a defined range was utilized for communications on the at least one communication path;

selectively updates at least one stored value of the network performance parameter based on the at least one delay parameter and an asymmetric weight if the bandwidth within the defined range was utilized for communications on the at least one communication path during a corresponding time interval, wherein the asymmetric weight is based on relative values of the at least one delay parameter for the at least one communications path over the corresponding time interval and at least one delay parameter associated with the at least one communications path over a time interval prior to the corresponding time interval; and

assesses network performance via comparison of a measured value of the network performance parameter to the selectively updated at least one stored value.

18. The networked computer of claim 17, wherein the assessment module further:

maintains an asymmetric moving average of the network performance parameter over the plurality of intervals.

19. The networked computer of claim 17, wherein determination of whether the bandwidth within the defined range was utilized for communications on the at least one communication path includes a determination of whether the at least one communications path is congested.

20. The networked computer of claim 17, wherein:

the selective updates to the at least one stored value of the network performance parameter are also based on extreme values during individual time intervals; and

the extreme value during a particular individual time interval of the plurality of time intervals comprises a maximum bandwidth during that particular time interval;

or

the extreme value during the particular individual time interval comprises a minimum bandwidth during that particular time interval.