Quantifying the impact of network latency on the end-to-end response time of distributed applications

• US 20100191505A1
• Filed: 01/23/2009
• Published: 07/29/2010
• Est. Priority Date: 01/23/2009
• Status: Active Grant

• Current Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Original Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Inventors: Sanders, William, Chen, Shuyi, Joshi, Kaustubb, Hiltunen, Matti, Schlichting, Richard
• Granted Patent: US 8,073,655 B2
• Time in Patent Office: Days
• US Class Current: 702/182
• CPC Class Codes: H04L 43/0852 Delays
• CPC Class Codes: H04L 43/10 Active monitoring, e.g. heartbeat, ping or trace-route

Abstract

A method for measuring system response sensitivity, using live traffic and an analysis that converts randomly arriving stimuli and reactions to the stimuli to mean measures over chosen intervals, thereby creating periodically occurring samples that are processed. The system is perturbed in a chosen location of the system in a manner that is periodic with frequency p, and the system'"'"'s response to arriving stimuli is measured at frequency p. The perturbation, illustratively, is with a square wave pattern.

Claims

1. A method in a distributed system where relationship between a property of a first component of said system and an end-to-end metric of said system is determined by injecting a failure in said first component and identifying effect of said injecting on said metric, the improvement comprising

replacing said injecting of a failure with injecting of a change in said component that is allows said distributed system to continue operating at normal mode;

and

quantitatively evaluating said relationship.

2. The method of claim 1 where said relationship is sensitivity of said end-to-end metric to said change in said first component.

3. The method of claim 1 where said change is a perturbation in of an attribute of said first component that affects operation of said system.

4. The method of claim 1 where said change is a change in delay of response of said first component to an applied signal.

5. The method of claim 1 where said measuring operates on signals that pass through said system in normal course of the system'"'"'s operation.

6. The method of claim 1 where said metric is mean processing delay of said system before a response is submitted to a processing request.

7. The method of claim 6 where said first component is a link of said system.

8. The method of claim 6 where in said step of quantitatively evaluating said mean processing delay the mean is over intervals Δ

T_s and said quantitatively evaluating is over a frame of N Δ

T_s intervals.

9. The method of claim 8 where said quantitatively evaluating is over a frame of N Δ

T_s intervals, carried out at a frequency p, where p<

N.

10. The method of claim 8 where said quantitatively evaluating is over a frame of N Δ

T_s intervals, carried out at a frequency p, where p=N/2k_d, and k_d is an integer.

11. The method of claim 1 where said change is periodic, at frequency p, and said step of quantitatively evaluating is carried out over a frame of N samples, and evaluated at frequency p, where p=N/2k_d, and k_d is an integer.

12. The method of claim 1 where said injected change has a square wave pattern.

13. The method of claim 1 where said metric is requests of a particular type that are presented to said system.

14. The method of claim 1 where said change is a change in clock frequency of a processor within said system.

15. The method of claim 1 further comprising the step of evaluating response time of said system when configuration of said system is changed to result in a change in latency of a link of said system.

16. The method of claim 15 where said step of evaluating is repeated a number of times, and said method further comprising changing said configuration of said system to optimize system response time.

17. The method of claim 1 where said steps of replacing and quantitatively evaluating is repeated a number of times, each time having a different component of said system having said injected change.

18. The method of claim 17 further comprising a step of choosing a component of said system for modifying in order to optimize response time of said system.

19. The method of claim 18 where said component is a processor of said system, and said modifying is a modification of a code module executed by said processor.

20. The method of claim 18 where said component is a processor of said system, and said modifying is a change in geographical location of said processor.

21. A method for measuring responsiveness of a system that receives requests at non-regular intervals, comprising the steps of:

undertaking a training time interval during which (a) intervals between arriving requests are noted, and (b) a statistical analysis is performed to identify an interval Δ

T_s within which there is a selected probability in excess of a chosen value that at least one request will arrive; and

measuring an attribute that represents average responsiveness of said system to requests arriving during each Δ

T_s interval.

22. The method of claim 21 where said requests are requests submitted to said system in course of said system'"'"'s normal operation.

23. The method of claim 21 further comprising the step perturbing operation of said system.

24. The method of claim 23 further comprising the step of:

calculating frequency spectrum of said measured attribute of a frame having N of said Δ

T_s intervals; and

said perturbing is for a duration of k_d of said Δ

T_s intervals, where k_d is an integer less than N.

25. The method of claim 24 where said measuring an attribute is evaluated for a frequency that is related to 1/k_d.

26. The method of claim 23 further comprising the step of:

calculating frequency spectrum of said measured attribute of a frame having N of said Δ

T_s intervals; and

said perturbing is periodic, with period 2k_d, where 2k_d≦

N.

27. The method of claim 26 where k_d=N/2p and p is an integer.

29. The method of claim 26 where said perturbing has a square wave pattern having a chosen magnitude A_d for k_d of said Δ

T_s intervals in each period of said square wave pattern.

30. The method of claim 23 where said perturbing alters an attribute of a chosen element of said system that affects responsiveness of said system.

31. The method of claim 30 where said chosen element is a link that interconnects two devices of said system, and said attribute is delay through said link.

32. The method of claim 30 where said chosen element is a processor of said system, and said attribute is processing speed of said system.

33. The method of claim 26 where said measuring an attribute is evaluated for a frequency that is related to 1/k_d.

34. The method of claim 21 where said system comprising hardware device and at least some of said devices are geographically distributed and interconnected via a network that carries traffic of other than said system.

28. The method of 26 where an attribute is evaluated for a frequency that is related to k_d.