Methods and apparatus for identifying chronic performance problems on data networks
View Patent ↗A system for identifying chronic performance problems on data networks includes network monitoring devices that provide measurements of performance metrics such as latency, jitter, and throughput, and a processing system for analyzing measurement data to identify the onset of chronic performance problems. Network behavior is analyzed to develop a baseline performance level (e.g., computing a mean and variance of a baseline sample of performance metric measurements). An operating performance level is measured during operation of the network (e.g., computing a mean and variance of an operating sample of performance metric measurements). A chronic performance problem is identified if the operating performance level exceeds a performance threshold and a difference between the baseline performance level and the operating performance level is determined to be statistically significant. A t-test based on the baseline mean and variance and the operating mean and variance can be used to determine if this difference is significant.
1. A method of identifying a chronic performance problem on a data network, comprising:
establishing a baseline performance level of at least one performance metric using a processor;
determining an operating performance level of the at least one performance metric during operation of the network using the processor;
identifying a chronic performance problem by performing both a first test and a second test, using the processor, the first test comprising a threshold test to determine whether the operating performance exceeds a performance threshold, the second test comprising a statistical test to determine whether a difference between a baseline performance level and the operating performance level is statistically significant, wherein the combination of the first and second tests distinguishes between chronic performance problems and transient performance problems and a chronic performance problem is identified in response to both the operating performance level exceeding the performance threshold and the difference between the baseline performance level and the operating performance level being statistically significant.
2. The method of claim 1 , wherein establishing a baseline performance level includes:
collecting a baseline sample comprising a plurality of measurements of the at least one performance metric taken over a period of time; and
computing a baseline mean and variance of the baseline sample, the baseline mean and variance being indicative of the baseline performance level.
3. The method of claim 1 , wherein determining an operating performance level includes:
collecting an operating sample comprising a plurality of measurements of the at least one performance metric taken over a period of time; and
computing an operating mean and variance of the operating sample, the operating mean and variance being indicative of the operating performance level.
4. The method of claim 1 , wherein the difference between the baseline performance level and the operating performance level is determined to be statistically significant by performing a t-test based on the baseline mean and variance and the operating mean and variance.
5. The method of claim 4 , further comprising:
selecting a confidence level threshold associated with performing the t-test.
6. The method of claim 1 , wherein the performance threshold is selected as a function of the baseline performance level.
7. The method of claim 1 , wherein the performance threshold is selected as a function of a predetermined performance level representing unacceptable performance.
8. The method of claim 1 , wherein the at least one performance metric includes at least one of: latency, jitter, throughput, data delivery ratio, and TCP retries.
9. The method of claim 1 , wherein the performance metric is associated with at least one of: a circuit, a channel, a link, a site, an application, a user, and a device.
10. A system for identifying chronic performance problems on a data network, comprising:
a plurality of monitoring devices configured to measure at least one performance metric at selected points in the network; and
a processor configured to receive performance metric measurements from the monitoring devices, wherein the processor determines a baseline performance level of the at least one performance metric, determines an operating performance level of the at least one performance metric during operation of the network, and identifies a chronic performance problem by performing both a first test and a second test, the first test comprising a threshold test to determine whether the operating performance exceeds a performance threshold, the second test comprising a statistical test to determine whether a difference between a baseline performance level and the operating performance level is statistically significant, wherein the combination of the first and second tests distinguishes between chronic performance problems and transient performance problems and the processor identifies a chronic performance problem in response to both the operating performance level exceeding the performance threshold and the difference between the baseline performance level and the operating performance level being statistically significant.
11. The system of claim 10 , wherein the processor receives from the plurality of monitoring devices a baseline sample comprising a plurality of measurements of the at least one performance metric taken over a period of time, and computes a baseline mean and variance of the baseline sample, the baseline mean and variance being indicative of the baseline performance level.
12. The system of claim 11 , wherein the processor receives from the plurality of monitoring devices an operating sample comprising a plurality of measurements of the at least one performance metric taken over a period of time, and computes an operating mean and variance of the operating sample, the operating mean and variance being indicative of the operating performance level.
13. The system of claim 10 , wherein the processor determines the difference between the baseline performance level and the operating performance level to be statistically significant by performing a t-test based on the baseline mean and variance and the operating mean and variance.
14. The system of claim 13 , wherein the processor performs the t-test based on a selectable confidence level threshold.
15. The system of claim 10 , wherein the processor selects the performance threshold as a function of the baseline performance level.
16. The system of claim 10 , wherein the processor selects the performance threshold as a function of a predetermined performance level representing unacceptable performance.
17. The system of claim 10 , wherein the at least one performance metric includes at least one of: latency, jitter, throughput, data delivery ratio, and TCP retries.
18. The system of claim 10 , wherein the performance metric is associated with at least one of: a circuit, a channel, a link, a site, an application, a user, and a device.
19. A system for identifying chronic performance problems on a data network, comprising:
means for measuring at least one performance metric at selected points in the network; and
means for determining a baseline performance level of the at least one performance metric;
means for determining an operating performance level of the at least one performance metric during operation of the network; and
means for identifying a chronic performance problem by performing both a first test and a second test, the first test comprising a threshold test to determine whether the operating performance exceeds a performance threshold, the second test comprising a statistical test to determine whether a difference between a baseline performance level and the operating performance level is statistically significant, wherein the combination of the first and second tests distinguishes between chronic performance problems and transient performance problems and a chronic performance problem is identified in response to the operating performance level exceeding the performance threshold and the difference between the baseline performance level and the operating performance level is statistically significant.