System and method for calibrating headroom margin

A computer readable medium storing instructions that, when executed by a computer, cause the computer to perform a method of determining a margin of capacity of a set of computing resources to be held in reserve, the method comprising:

• selecting a response time factor representing an acceptable level of response time for a workload, wherein the workload comprises a plurality of requests for a set of computing resources and wherein said response time factor reflects a magnitude of response time increase as utilization of the set of resources increases;
• from a first set of data indicating response times for different percentages of utilization of the set of resources, identifying a first utilization percentage corresponding to said selected response time factor;
• and
• selecting an initial margin based on said first identified utilization percentage.

A computer-implemented method of dynamically adjusting a margin of reserved capacity of a set of computing resources, the method comprising:

• operating a set of resources having a dynamically adjustable membership;
• selecting a response time increase factor representing a target maximum response time increase to be tolerated as the workload of the set of resources increases;
• from graph data plotting response time increase factors against percentages of utilization of the set of resources, identifying a first utilization percentage corresponding to said selected response time increase factor;
• setting the margin of capacity of the set of resources to be held in reserve to one hundred percent minus said first utilization percentage;
• collecting response time data for a first workload comprising varying levels of loads on the set of resources;
• from said collected response time data, assembling adjusted graph data plotting response time increase factors against percentages of utilization of the set of resources;
• from said adjusted graph data, identifying a second utilization percentage corresponding to said selected response time increase factor;
• and
• adjusting said margin based on said second utilization percentage.

A computer readable medium storing instructions that, when executed by a computer, cause the computer to perform a method of dynamically adjusting a margin of reserved capacity of a set of computing resources, the method comprising:

• operating a set of resources having a dynamically adjustable membership;
• selecting a response time increase factor representing a target maximum response time increase to be tolerated as the workload of the set of resources increases;
• from graph data plotting response time increase factors against percentages of utilization of the set of resources, identifying a first utilization percentage corresponding to said selected response time increase factor;
• setting the margin of capacity of the set of resources to be held in reserve to one hundred percent minus said first utilization percentage;
• collecting response time data for a first workload comprising varying levels of loads on the set of resources;
• from said collected response time data and said first workload, assembling adjusted graph data plotting response time increase factors against percentages of utilization of the set of resources;
• from said adjusted graph data, identifying a second utilization percentage corresponding to said selected response time increase factor;
• and
• adjusting said margin based on said second utilization percentage.

An apparatus for dynamically modifying a margin of reserved capacity of a set of resource, comprising:

• a set of computing resources for which requests are received;
• a first set of data plotting response time increase factors for varying percentages of utilization of the set of resources for randomly received requests;
• a collection of observed response times for different load levels placed on the set of resources;
• a second set of data plotting response time increase factors for varying percentages of utilization of the set of resources for randomly received requests, wherein said second set of data is derived from said collection of observed response times;
• and
• a processor configured to:
• analyze the first set of data to select an initial maximum usable percentage of the set of resources based on a specified acceptable response time increase factor;
• and
• analyze the second set of data to dynamically select an adjusted maximum usable percentage of the set of resources based on the specified acceptable response time increase factor;
• wherein an initial margin is set to one hundred percent minus said initial maximum usable percentage;
• and
• a dynamically adjusted margin is set to one hundred percent minus said adjusted maximum usable percentage.