Selecting resource allocation policies and resolving resource conflicts

Patent metadata

Field	Value
Publication number	US-9697045-B2
Application number	US-201514667374-A
Country	US
Kind code	B2
Filing date	Mar 24, 2015
Priority date	Mar 24, 2015
Publication date	Jul 4, 2017
Grant date	Jul 4, 2017

Abstract

Techniques for workload management in cloud computing infrastructures are provided. In one aspect, a method for allocating computing resources in a datacenter cluster is provided. The method includes the steps of: creating multiple, parallel schedulers; and automatically selecting a resource allocation method for each of the schedulers based on one or more of a workload profile, user requirements, and a state of the datacenter cluster, wherein an optimistic resource allocation method is selected for at least a first one or more of the schedulers and a pessimistic resource allocation method is selected for at least a second one or more of the schedulers. Due to optimistic resource allocation conflicts may arise. Methods to resolve such conflicts are also provided.

First claim

What is claimed is:

1. A method for allocating computing resources in a datacenter cluster, the method comprising steps of:
creating multiple, parallel schedulers;
automatically selecting a resource allocation method for each of the schedulers based on one or more of a workload profile, user requirements, and a state of the datacenter cluster, wherein an optimistic resource allocation method is selected for at least a first one or more of the schedulers and a pessimistic resource allocation method is selected for at least a second one or more of the schedulers;
receiving new workloads;
classifying the new workloads; and
based on the classifying of the new workloads, determining whether to i) assign the new workloads to an existing one of the schedulers or ii) create one or more new schedulers and assign the new workloads to the one or more new schedulers, and any interferences among the new workloads are resolved by observing one or more of a number of cancelled decisions, and datacenter cluster state from previous deployments so as to relieve cancellations and thus maintain a higher workload performance.

2. The method of claim 1 , wherein the pessimistic resource allocation method comprises one or more of full pre-allocation of the computing resources and partial pre-allocation of the computing resources, wherein the one or more of full pre-allocation and partial pre-allocation comprises blocking the computing resources.

3. The method of claim 1 , wherein the optimistic resource allocation method comprises non-pre-allocation of the computing resources.

4. The method of claim 1 , wherein the datacenter cluster comprises a cloud computing infrastructure.

5. The method of claim 4 , wherein the cloud computing infrastructure comprises multiple clouds.

6. The method of claim 1 , the steps further comprising a step of:
dynamically switching the resource allocation method of one or more of the schedulers from the optimistic resource allocation method to the pessimistic resource allocation method.

7. The method of claim 6 , wherein the resource allocation method of the one or more of the schedulers is dynamically switched from the optimistic resource allocation method to the pessimistic resource allocation method either manually or automatically based on a user's decision.

8. The method of claim 6 , wherein the resource allocation method of the one or more of the schedulers is dynamically switched from the optimistic resource allocation method to the pessimistic resource allocation method either manually or automatically based on a change in the state of the datacenter cluster.

9. The method of claim 1 , wherein the new workloads are classified based on one or more of central processing unit (CPU), memory, network, and disk intensiveness.

10. The method of claim 1 , wherein the new workloads are classified based on data from one or more previous iterations of the method.

11. The method of claim 1 , the steps further comprising a step of:
identifying conflicts in scheduler decisions.

12. The method of claim 11 , wherein the conflicts comprise one or more of workload interference conflicts, server overloading conflicts, and missing resource conflicts.

13. The method of claim 11 , the steps further comprising a step of:
selecting resolution policies for the conflicts.

14. The method of claim 13 , wherein the resolution policies for the conflicts are selected based on data from one or more previous iterations of the method.

15. A method for allocating computing resources in a datacenter cluster, the method comprising steps of:
creating multiple, parallel schedulers;
automatically selecting a resource allocation method for each of the schedulers based on one or more of a workload profile, user requirements, and a state of the datacenter cluster, wherein an optimistic resource allocation method is selected for at least a first one or more of the schedulers and a pessimistic resource allocation method is selected for at least a second one or more of the schedulers;
receiving new workloads;
classifying the new workloads;
based on the classifying of the new workloads, determining whether to i) assign the new workloads to an existing one of the schedulers or ii) create one or more new schedulers and assign the new workloads to the one or more new schedulers;
identifying conflicts in scheduler decisions; and
selecting resolution policies for the conflicts by observing one or more of a number of cancelled decisions, and datacenter cluster state from previous deployments so as to relieve cancellations and thus maintain a higher workload performance.

16. The method of claim 15 , wherein the datacenter cluster comprises a cloud computing infrastructure.

17. The method of claim 16 , wherein the cloud computing infrastructure comprises multiple clouds.

18. A computer program product for allocating computing resources in a datacenter cluster, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to:
create multiple, parallel schedulers;
automatically select a resource allocation method for each of the schedulers based on one or more of a workload profile, user requirements, and a state of the datacenter cluster, wherein an optimistic resource allocation method is selected for at least a first one or more of the schedulers and a pessimistic resource allocation method is selected for at least a second one or more of the schedulers;
receive new workloads;
classify the new workloads;
based on the classification of the new workloads, determine whether to i) assign the new workloads to an existing one of the schedulers or ii) create one or more new schedulers and assign the new workloads to the one or more new schedulers;
identify conflicts in scheduler decisions; and
select resolution policies for the conflicts by observing one or more of a number of cancelled decisions, and datacenter cluster state from previous deployments so as to relieve cancellations and thus maintain a higher workload performance.

Assignees

• IBM
• Barcelona Supercomputing Center—Centro Nac De Supercomputación

Inventors

• Amaral Marcelo
• Doganata Yurdaer N
• Mohomed Iqbal I
• Tantawi Asser N
• Unuvar Merve

Classifications

• G06F9/5027 Primary

  the resource being a machine, e.g. CPUs, Servers, Terminals · CPC title

• G06F9/5005 Primary

  to service a request · CPC title

• G06F9/5088

  involving task migration · CPC title

• G06F9/4881 Primary

  Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues · CPC title