Evaluating and predicting computer system performance using kneepoint analysis
AssigneeNetApp, Inc.
InventorsGeorge Smirnov, Leon Fairbanks, Kenneth Hu, David Kaeli
Life of this patent
Estimating the term…
Abstract
The present invention provides methods, systems, apparatus, and computer software/program code products adapted for operating in, or in conjunction with, an otherwise conventional computing system, and which enable evaluating, monitoring and predicting the performance of computer systems and individual elements or groups of elements within such computer systems.
Claims19
A method of determining whether the performance of a computing system is likely to scale exponentially or faster than exponentially with respect to increased system load, the method comprising:
- measuring utilization of one or more computing system components or groups of components to generate utilization information, wherein the utilization may vary over time;
- generating, using the utilization information, an operating curve characteristics of the computing system;
- determining at least one utilization value on the operating curve to be an x-coordinate of a kneepoint of the operating curve, the kneepoint being a unique property of the operating curve and being substantially constant for a performance queuing model representing one or more aspects of the performance of the computing system;
- comparing a given utilization value to the x-coordinate of the kneepoint;
- setting a bound based on the x-coordinate of the kneepoint;
- and
- indicating an alarm condition when the given utilization value is above the bound based on the x-coordinate of the kneepoint.
A system for determining whether the performance of a computing system is likely to scale exponentially or faster than exponentially with respect to increased system load, the system comprising:
- means for measuring utilization of one or more computing system components or groups of components to generate utilization information, wherein the utilization may vary over time;
- means for generating, using the utilization information, an operating curve characteristics of the computing system;
- means for determining at least one utilization value on the operating curve to be an x-coordinate of a kneepoint of the operating curve, the kneepoint being a unique property of the operating curve and being substantially constant for a performance queuing model representing one or more aspects of the performance of the computing system;
- means for comparing a given utilization value to the x-coordinate of the kneepoint;
- means for setting a bound based on the x-coordinate of the kneepoint;
- and
- means for indicating an alarm condition when the given utilization value is above the bound based on the x-coordinate of the kneepoint.
A computer program product comprising comprising computer-readable program instructions encoded on a computer-readable medium and executable by a computer to enable the computer to determine whether the performance of a computing system is likely to scale exponentially or faster than exponentially with respect to increased system load, the computer program product comprising:
- A. first computer-readable program instructions, encoded on the computer-readable medium and executable by the computer to enable the computer to measure utilization of one or more computing system components or groups of components to generate utilization information, wherein the utilization may vary over time;
- B. second computer-readable program instructions, encoded on the computer-readable medium and executable by the computer to enable the computer to generate, using the utilization information, an operating curve characteristics of the computing system;
- C. third computer-readable program instructions, encoded on the computer-readable medium and executable by the computer to enable the computer to determine at least one utilization value on the operating curve to be an x-coordinate of a kneepoint of the operating curve, the kneepoint being a unique property of the operating curve and being substantially constant for a performance queuing model representing one or more aspects of the performance of the computing system;
- D. fourth computer-readable program instructions, encoded on the computer-readable medium and executable by the computer to enable the computer to compare a given utilization value to the x-coordinate of the kneepoint;
- E. fifth computer-readable program instructions, encoded on the computer-readable medium and executable by the computer to enable the computer to set a bound based on the x-coordinate of the kneepoint;
- and
- F. sixth computer-readable program instructions, encoded on the computer-readable medium and executable by the computer to enable the computer to indicate an alarm condition when the given utilization value is above the bound based on the x-coordinate of the kneepoint.
In a computing system, an improvement for determining whether the performance of a computing system is likely to scale exponentially or faster than exponentially with respect to increased system load, the improvement comprising:
- means for measuring utilization of one or more computing system components or groups of components to generate utilization information, wherein the utilization may vary over time;
- means for generating, using the utilization information, an operating curve characteristics of the computing system;
- means for determining at least one utilization value on the operating curve to be an x-coordinate of a kneepoint of the operating curve, the kneepoint being a unique property of the operating curve and being substantially constant for a performance queuing model representing one or more aspects of the performance of the computing system;
- means for comparing a given utilization value to the x-coordinate of the kneepoint;
- means for setting a bound based on the x-coordinate of the kneepoint;
- and
- means for indicating an alarm condition when the given utilization value is above the bound based on the x-coordinate of the kneepoint.
Dependencies are inferred from the claim text.