Shared resource contention

Patent metadata

Field	Value
Publication number	US-9954757-B2
Application number	US-201314759659-A
Country	US
Kind code	B2
Filing date	Jan 23, 2013
Priority date	Jan 23, 2013
Publication date	Apr 24, 2018
Grant date	Apr 24, 2018

Abstract

Contention for shared resources in a shared resource environment may be determined based on measurements from a probe running in the shared resource environment. The measurements can be compared to benchmarks, and a contention value may be determined based on the comparison.

First claim

What is claimed is:

1. A shared resource environment contention determination system comprising:
a load generator executed by a processor to determine measurements from a probe, simulating a resource consumer, running with virtual machines (VMs) in a contention region in a shared resource environment, wherein the measurements measure performance of shared resources utilized by the probe and the VMs in the contention region, compare the measurements to benchmarks, and determine a contention value representative of an amount of contention among the VMs running in the contention region based on the comparison, wherein the contention comprises interference among the VMs utilizing the shared resources,
wherein an available capacity of a shared resource in the contention region is calculated based on the contention value,
wherein the available capacity comprises a number of shares of one of the shared resources that are available for allocation to a VM in the contention region, and the available capacity A is equal to C−f(P+N), wherein C is total capacity in terms of shares of the shared resource, f is the contention value, P is a number of shares already provisioned and N is a number of shares required for an additional VM in the contention region.

2. The system of claim 1 , wherein the shared resources in the contention region include a virtual machine monitor and an L3 cache.

3. The system of claim 2 , wherein the measurements for the shared resources comprise connection arrival rates measured for the virtual machine monitor and cache misses measured for the L3cache.

4. The system of claim 1 , wherein the contention causes degradation in performance of applications running on the VMs that is determined from the contention value.

5. The system of claim 1 , wherein the probe comprises code executed in the contention region to simulate an operation of an application running on a VM in the contention region.

6. The system of claim 1 , wherein the measurements are for metrics measuring the performance of the shared resources in the contention region, and the benchmarks comprise measurements for the metrics measured by the probe running in the contention region in isolation without sharing the shared resources with the VMs.

7. The system of claim 1 , wherein the probe is tuned based on a type of the shared resource being measured and attributes of the shared resource.

8. The system of claim 1 , wherein the load generator is to determine if the amount of contention exceeds a threshold and to report the contention value to a management system to control the VMs if the amount of contention exceeds the threshold.

9. The system of claim 8 , wherein the management system is to migrate a VM to reduce the impact of contention on the VMs in response to the amount of contention exceeding the threshold.

10. A method of estimating contention among resource consumers in a contention region in a shared resource environment, the method comprising:
determining measurements of performance of shared resources in the contention region from a probe running with the resource consumers and simulating a workload on the shared resources;
comparing the measurements to benchmarks; and
determining, by a processor, a contention value representative of an amount of contention among the resource consumers running in the contention region based on the comparison,
wherein the contention comprises interference among the resource consumers and the probe utilizing shared resources in the contention region and the contention causes a degradation in performance of the resource consumers,
wherein an available capacity of a shared resource in the contention region is calculated based on the contention value,
wherein the available capacity comprises a number of shares of one of the shared resources that are available for allocation to a VM in the contention region, and the available capacity A is equal to C−f(P+N), wherein C is total capacity in terms of shares of the shared resource, f is the contention value, P is a number of shares already provisioned and N is a number of shares required for an additional VM in the contention region.

11. The method of claim 10 , comprising:
determining if the amount of contention exceeds a threshold; and
sending the contention value to a management system if the amount of contention exceeds the threshold to control consumption of the shared resources by the resource consumers based on the contention value.

12. The method of claim 10 , wherein the resource consumers comprise VMs running applications or applications not running on a VM.

13. The method of claim 10 , wherein the probe comprises code executed in the contention region to simulate an operation of an application running on a VM in the contention region.

14. The method of claim 10 , wherein the contention causes degradation in performance of a plurality of applications running on a plurality of VMs that is determined from the contention value.

15. The method of claim 10 , wherein determining measurements of performance of a virtual machine monitor and an L3 cache in a contention region further comprises determining if the amount of contention exceeds a threshold and to report the contention value to a management system to control a plurality of VMs if the amount of contention exceeds the threshold.

16. A non-transitory computer readable medium including machine readable instructions executable by at least one processor to:
determine measurements of performance of a virtual machine monitor and an L3 cache in a contention region
from a probe running with VMs in the contention region, wherein the probe simulates a VM running in the contention region;
compare the measurements to benchmarks; and
determine a contention value representative of an amount of contention among the VMs running in the contention region based on the comparison, wherein the contention comprises interference among the VMs using the virtual machine monitor and the L3 cache;
wherein an available capacity of a shared resource in the contention region is calculated based on the contention value; and
wherein the available capacity comprises a number of shares of one of the shared resources that are available for allocation to a VM in the contention region, and the available capacity A is equal to C−f(P+N), wherein C is total capacity in terms of shares of the shared resource, f is the contention value, P is a number of shares already provisioned and N is a number of shares required for an additional VM in the contention region.

17. The non-transitory computer readable medium of claim 16 , wherein the probe comprises code executed in the contention region to simulate an operation of an application running on a VM in the contention region.

18. The non-transitory computer readable medium of claim 16 , wherein the probe is tuned based on a type of the shared resource being measured and attributes of the shared resource.

Assignees

• Hewlett Packard Development Co
• Hewlett Packard Entpr Dev Lp

Inventors

• Hyser Christopher D
• Rolia Jerome
• Krishnamurthy Diwakar
• Mukherjee Joydeep

Classifications

• G06F2009/45591

  Monitoring or debugging support · CPC title

• G06F11/00

  Error detection; Error correction; Monitoring (error detection, correction or monitoring in information storage based on relative movement between record carrier and transducer G11B20/18; monitoring, i.e. supervising the progress of recording or reproducing G11B27/36; in static stores G11C29/00) · CPC title

• H04L43/0876

  Network utilisation, e.g. volume of load or congestion level · CPC title

• G06F9/5088 Primary

  involving task migration · CPC title

• G06F2209/508

  Monitor · CPC title