Capacity over-commit management in resource provisioning environments

• US 9,245,246 B2
• Filed: 09/13/2012
• Issued: 01/26/2016
• Est. Priority Date: 04/22/2010
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Breitgand, David, Epstein, Amir
• Primary Examiner(s): Bates, Kevin
• Assistant Examiner(s): Scott, Mark A
• Application Number:
  US13/612,855
• Publication Number:
  US 20130007272A1
• Time in Patent Office: 1,230 Days
• Field of Search: 709/224
• US Class Current: 1/1
• CPC Class Codes: G06F 9/5077 Logical partitioning of resources; Management or configuration of virtualized resources specific details on emulation or internal functioning of virtual machines G06F9/455
• CPC Class Codes: G06Q 10/06 Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling

Abstract

Systems and methods for managing capacity over-commitment in a resource provisioning environment. The method comprising collecting historical data about capacity usage in a resource provisioning environment and failure statistics from an operational database, in addition to prior service commitments due to previously contracted service level agreements (SLAs) stored in a SLA repository; calculating equivalent capacity for the resource provisioning environment based on a defined residual benefit goal, and other business goals instrumented by way of a policy engine; and determining whether placing requested virtual resources on one or more hosts is feasible taking into account anti-collocation constraints that prohibit placement of two or more virtual resources on same host.

Claims

1. A computer-implemented method for managing capacity over-commitment in a resource provisioning environment, the method comprising:

collecting historical data about capacity usage in a resource provisioning environment and failure statistics from an operational database, in addition to prior service commitments due to previously contracted service level agreements (SLAs) stored in a SLA repository;

calculating equivalent capacity for the resource provisioning environment based on a defined residual benefit goal, and other business goals instrumented by way of a policy engine;

determining whether placing requested virtual resources on one or more hosts is feasible, in response to determining that sufficient physical capacity is available to host the calculated equivalent capacity and taking into account anti-collocation constraints that prohibit placement of two or more virtual resources on same host, wherein the determining is repeated a number of times (N), wherein the number of times (N) is automatically computed by estimating a confidence interval for a mean of Normal distribution based on a defined significance level, and wherein the determining whether placing requested virtual resources on one or more hosts is feasible comprises checking a target risk level of resource congestion does not exceed a threshold risk probability based on determining a percentage of failures during the repeated number of (N) times; and

issuing an unsafe over-commit level warning based on the determination that placing requested virtual resources on one or more hosts is not feasible.

2. The method of claim 1 wherein taking into account anti-collocation constraints comprises:

computing empirical probability distribution functions (PDF) for virtual resources in a service cluster; and

generating a random placement problem instance based on the computed empirical PDFs.

3. The method of claim 2 wherein the empirical PDFs are computed based on predetermined acceptable risk levels defined for a requested service.

4. The method of claim 1 further comprising admitting a requested service by the service subscriber, in response to determining that placing requested virtual resources on one or more hosts is feasible such that sufficient physical capacity is available to host the calculated equivalent capacity.

5. The method of claim 4 further comprising generating a placement problem instance and validating the placement.

6. The method of claim 5 further comprising authorizing capacity over-commitment, in response to determining that risk of failure as a result of over-commitment is below a first threshold.

7. The method of claim 1 further comprising notifying a capacity planner that additional physical capacity is needed to cover the calculated equivalent capacity, in response to determining that the physical capacity is insufficient to host the calculated equivalent capacity.

8. A computer-implemented system for managing capacity over-commitment in a resource provisioning environment, the system comprising:

one or more processors to executed instruction loaded from a memory;

a logic unit for collecting historical data about capacity usage in a resource provisioning environment and failure statistics from an operational database, in addition to prior service commitments due to previously contracted service level agreements (SLAs) stored in a SLA repository;

a logic unit for calculating equivalent capacity for the resource provisioning environment based on a defined residual benefit goal, and other business goals instrumented by way of a policy engine;

a logic unit for determining whether placing requested virtual resources on one or more hosts is feasible, in response to determining that sufficient physical capacity is available to host the calculated equivalent capacity and taking into account anti-collocation constraints that prohibit placement of two or more virtual resources on same host, wherein the determining is repeated a number of times (N), wherein the number of times (N) is automatically computed by estimating a confidence interval for a mean of Normal distribution based on a defined significance level, and wherein the determining whether placing requested virtual resources on one or more hosts is feasible comprises checking a target risk level of resource congestion does not exceed a threshold risk probability based on determining a percentage of failures during the repeated number of (N) times; and

issuing an unsafe over-commit level warning based on the determination that placing requested virtual resources on one or more hosts is not feasible.

9. The system of claim 8 wherein the anti-collocation constraints are taking into account by computing empirical probability distribution functions (PDF) for virtual resources in a service cluster;

and generating a random placement problem instance based on the computed empirical PDFs.

10. The system of claim 8 wherein the empirical PDFs are computed based on predetermined acceptable risk levels defined for a requested service.

11. A computer program product comprising a non-transitory computer readable storage medium having a computer readable program, wherein the computer readable program when executed on a computer causes the computer to:

collect historical data about capacity usage in a resource provisioning environment and failure statistics from an operational database, in addition to prior service commitments due to previously contracted service level agreements (SLAs) stored in a SLA repository;

calculate equivalent capacity for the resource provisioning environment based on a defined residual benefit goal, and other business goals instrumented by way of a policy engine;

determine whether placing requested virtual resources on one or more hosts is feasible, in response to determining that sufficient physical capacity is available to host the calculated equivalent capacity and taking into account anti-collocation constraints that prohibit placement of two or more virtual resources on same host, wherein the determining is repeated a number of times (N), wherein the number of times (N) is automatically computed by estimating a confidence interval for a mean of Normal distribution based on a defined significance level, and wherein the determining whether placing requested virtual resources on one or more hosts is feasible comprises checking a target risk level of resource congestion does not exceed a threshold risk probability based on determining a percentage of failures during the repeated number of (N) times; and

issuing an unsafe over-commit level warning based on the determination that placing requested virtual resources on one or more hosts is not feasible.

12. The computer program product of claim 11 wherein account anti-collocation constraints are taken into account by computing empirical probability distribution functions (PDF) for virtual resources in a service cluster;

and generating a random placement problem instance based on the computed empirical PDFs.

13. The computer program product of claim 11 wherein the empirical PDFs are computed based on predetermined acceptable risk levels defined for a requested service.