Automatic serial order starting of resource groups on failover systems based on resource group usage prediction

Patent metadata

Field	Value
Publication number	US-9934098-B2
Application number	US-201514851150-A
Country	US
Kind code	B2
Filing date	Sep 11, 2015
Priority date	Jul 21, 2015
Publication date	Apr 3, 2018
Grant date	Apr 3, 2018

Abstract

A method for automatically prioritizing startup of resource groups during a migration event. The method may include monitoring resource usage of a first and a second set of applications associated, respectively, with a first and a second resource group executing on a first computing node. The method may additionally include generating respective first and second resource usage models for the first and second resource groups based on resource usage. The method may then include extrapolating, based on the first and second resource usage models, respective first and second resource group usage scores for the first and second resource groups at a second time in response to a migration event, the second time occurring subsequent to the first time. The method may further include determining, based on the extrapolating, a priority order for serially starting the first and second set of applications on a second computing node at the second time.

First claim

What is claimed is:

1. A method comprising:
monitoring system resource usage of a first set of applications in a first resource group of a set of at least two resource groups executing on a first computing node at a first time,
wherein each resource group of the set of at least two resource groups comprises one or more related applications that execute on the first computing node and utilize system resources of the first computing node;
monitoring system resource usage of a second set of applications in a second resource group of the set of at least two resource groups executing on the first computing node at the first time;
generating, based on the monitored system resource usage of the first set of applications, a first resource usage model, wherein the first resource usage model comprises, for each application in the first set of applications, a first resource usage score which aggregates monitored system resource usage of two or more system resources used by the each application in the first set of applications;
generating, based on the monitored resource usage of the second set of applications, a second resource usage model, wherein the second resource usage model comprises, for each application in the second set of applications, a second resource usage score which aggregates monitored system resource usage of two or more system resources used by the each application in the second set of applications;
generating, based on the first resource usage model and in response to an occurrence of a migration event, a first resource group usage score for the first resource group, wherein,
the migration event is an event requiring each resource group of the set of at least two resource groups to be moved from executing on the first computing node to executing on a second computing node, and
the first resource group usage score is a prediction of aggregate system resource usage of applications in the first resource group subsequent to the occurrence of the migration event;
generating, based on the second resource usage model and in response to the occurrence of the migration event, a second resource group usage score for the second resource group, wherein the second resource group usage score is a prediction of aggregate system resource usage of applications in the second resource group subsequent to the occurrence of the migration event;
determining, based on at least the first resource group usage score and second resource group usage score, that the first resource group is predicted to have greater likelihood of being in use after the migration event than the second resource group;
generating, in response to the determining, a priority order for serially loading the first resource group and second resource group on a second computing node; and
serially loading based on the priority order, the first resource group on the second computing node before the second resource group.

2. The method of claim 1 , wherein generating the first resource usage model comprises:
determining, based on the monitored system resource usage of the first set of applications, a resource usage score for a first application in the first set of applications by:
determining a first resource usage value that indicates a quantity of a first system resource used by the first application,
determining a second resource usage value that indicates quantity of a second system resource used by the first application,
determining a first weight and a second weight for additively combining, respectively, the first resource usage value and the second resource usage value in a first mathematical expression, wherein the first weight and the second weight respectively indicate a relative importance, with respect to all system resources used by the first application, of the first system resource and the second system resource for determining the resource usage score for the first application, and
evaluating the first mathematical expression to determine the resource usage score for the first application.

3. The method of claim 2 , wherein generating the first resource usage model comprises:
storing, for each application in the first set of applications, the first resource usage score in a usage score history; and
generating the first resource usage model using the usage score history.

4. The method of claim 2 , further comprising:
determining, based on the monitored system resource usage of the first set of applications, a resource usage score for a second application of the first set of applications by
determining a third resource usage value that indicates a quantity of a third system resource used by the second application,
determining a fourth resource usage value that indicates quantity of a fourth system resource used by the second application,
determining a third weight and a fourth weight to additively combine, respectively, the third resource usage value and the fourth resource usage value in a second mathematical expression, wherein the third weight and the fourth weight respectively indicate a relative importance, with respect to all system resources used by the second application, of the third system resource and the fourth system resource for determining the resource usage score for the second application, and
evaluate the second mathematical expression to determine the resource usage score for the second application;

determining a fifth weight and a sixth weight to additively combine, respectively, the resource usage score for the first application and the resource usage score for the second application in a third mathematical expression, wherein the fifth weight and the sixth weight respectively indicate a relative importance, with respect to all applications in the first set of applications, of the first application and the second application to determining the first resource group usage score; and
evaluating the third mathematical expression to determine the first resource group usage score.

5. The method of claim 1 , wherein the monitoring comprises measuring at least one of central processor usage, memory utilization, storage utilization, and network utilization.

6. The method of claim 1 , further comprising:
serially loading the first resource group by executing operations to execute each application in the first set of applications on the second computing node.

7. The method of claim 1 , wherein the first set of applications has a higher priority than the second set of applications when the first resource group usage score is larger than the second resource group usage score.

8. The method of claim 6 , further comprising:
adjusting, prior to the serially loading, the priority order according to a user defined priority list.

9. The method of claim 6 , further comprising:
adjusting, prior to the serially loading, the priority order according to a hierarchical relationship between resource groups.

10. The method of claim 1 , wherein the migration event is a failover event.

11. The method of claim 1 , wherein the first resource usage model and second resource usage model are stochastic models.

Assignees

• IBM

Inventors

• Bell Dionysius D
• Li Li
• Madduri Hari H
• Mcnair Ryan D

Classifications

• G06F2201/85

  Active fault masking without idle spares · CPC title

• G06F9/5077

  Logical partitioning of resources; Management or configuration of virtualized resources (specific details on emulation or internal functioning of virtual machines G06F9/455) · CPC title

• G06F11/1438 Primary

  Restarting or rejuvenating · CPC title

• G06F9/4862 Primary

  the task being a mobile agent, i.e. specifically designed to migrate · CPC title

• G06F11/2028

  eliminating a faulty processor or activating a spare · CPC title