Object-based load balancing approaches in distributed storage system

Patent metadata

Field	Value
Publication number	US-11507409-B2
Application number	US-202016820735-A
Country	US
Kind code	B2
Filing date	Mar 17, 2020
Priority date	Jan 22, 2020
Publication date	Nov 22, 2022
Grant date	Nov 22, 2022

Abstract

One example method to perform object-based load balancing in a distributed storage system of a virtualization system supported by a cluster of host machines may include determining, by a first host machine in the cluster, whether any host machine in the cluster has affinity to a site. The method may also include determining, by the first host machine, whether to distribute affinity Internet small computer system interface (iSCSI) targets owned by the first host machine to at least the second machine based on a first balance objective associated with the site, and after having considered the first balance objective, determining, by the first host machine, whether to distribute iSCSI targets without affinity owned by the first host machine to other host machines in the cluster based on a second balance objective associated with the cluster.

First claim

We claim:

1. A method to perform object-based load balancing in a distributed storage system of a virtualization system supported by a cluster of host machines, the method comprising:
determining, by a first host machine in the cluster, whether any host machine in the cluster has affinity to a site;
in response to the first host machine and at least a second host machine in the cluster having the affinity to the site, distributing, by the first host machine, affinity Internet small computer system interface (iSCSI) targets owned by the first host machine to at least the second host machine based on a first balance objective associated with the site; and
after having considered the first balance objective, distributing, by the first host machine, iSCSI targets without affinity owned by the first host machine to other host machines in the cluster based on a second balance objective associated with the cluster.

2. The method of claim 1 , further comprising:
calculating a first average weight for the site based on a first target weight associated with the affinity iSCSI targets in the site and a first number associated with any host machine in the cluster having affinity to the site; and
distributing the affinity iSCSI targets owned by the first host machine based on a comparison between the first average weight and a number of the affinity iSCSI targets owned by the first host machine.

3. The method of claim 2 , further comprising:
calculating a second average weight for the cluster based on a second target weight associated with the iSCSI targets without affinity in the cluster and a second number associated with all host machines in the cluster; and
distributing the iSCSI targets without affinity owned by the first host machine based on a comparison between the second average weight and a number of the iSCSI targets without affinity owned by the first host machine.

4. The method of claim 2 , further comprising:
in response to the number of the affinity iSCSI targets owned by the first host machine exceeding the first average weight, moving, by the first host machine, excessive affinity iSCSI targets from the first host machine to any of host machines in the site owning less number of affinity iSCSI targets than the first average weight based on the first balance objective.

5. The method of claim 3 , further comprising:
in response to the number of the iSCSI targets without affinity owned by the first host machine exceeding the second average weight, moving, by the first host machine, excessive iSCSI targets without affinity from the first host machine to any of the other host machine in the cluster owning less number of iSCSI targets without affinity than the second average weight based on the second balance objective.

6. The method of claim 1 , further comprising:
generating a first mapping, by the first host machine, based on information associated with each host machine in the cluster having the affinity to the site and a number of affinity iSCSI targets owned by the each host machine in the cluster having the affinity to the site.

7. The method of claim 1 , further comprising:
generating a second mapping, by the first host machine, based on information associated with each host machine in the cluster and a number of iSCSI targets without affinity owned by the each host machine in the cluster.

8. A non-transitory computer-readable storage medium that includes a set of instructions which, in response to execution by a processor of a first host machine in a cluster of host machines, cause the processor to perform operations of object-based load balancing in a distributed storage system of a virtualization system supported by the cluster of host machines, the operations comprising:
determining, by the first host machine, whether any host machine in the cluster has affinity to a site;
in response to the first host machine and at least a second host machine in the cluster having the affinity to the site, distributing, by the first host machine, affinity Internet small computer system interface (iSCSI) targets owned by the first host machine to at least the second host machine based on a first balance objective associated with the site; and
after having considered the first balance objective, distributing, by the first host machine, iSCSI targets without affinity owned by the first host machine to other host machines in the cluster based on a second balance objective associated with the cluster.

9. The non-transitory computer-readable medium of claim 8 , wherein the operations further comprise:
calculating a first average weight for the site based on a first target weight associated with the affinity iSCSI targets in the site and a first number associated with any host machine in the cluster having affinity to the site; and
distributing the affinity iSCSI targets owned by the first host machine based on a comparison between the first average weight and a number of the affinity iSCSI targets owned by the first host machine.

10. The non-transitory computer-readable medium of claim 9 , wherein the operations further comprise:
calculating a second average weight for the cluster based on a second target weight associated with the iSCSI targets without affinity in the cluster and a second number associated with all host machines in the cluster; and
distributing the iSCSI targets without affinity owned by the first host machine based on a comparison between the second average weight and a number of the iSCSI targets without affinity owned by the first host machine.

11. The non-transitory computer-readable medium of claim 9 , wherein the operations further comprise:
in response to the number of the affinity iSCSI targets owned by the first host machine exceeding the first average weight, moving, by the first host machine, excessive affinity iSCSI targets from the first host machine to any of host machines in the site owning less number of affinity iSCSI targets than the first average weight based on the first balance objective.

12. The non-transitory computer-readable medium of claim 10 , wherein the operations further comprise:
in response to the number of the iSCSI targets without affinity owned by the first host machine exceeding the second average weight, moving, by the first host machine, excessive iSCSI targets without affinity from the first host machine to any of the other host machine in the cluster owning less number of iSCSI targets without affinity than the second average weight based on the second balance objective.

13. The non-transitory computer-readable medium of claim 8 , wherein the operations further comprise:
generating a first mapping, by the first host machine, based on information associated with each host machine in the cluster having the affinity to the site and a number of affinity iSCSI targets owned by the each host machine in the cluster having the affinity to the site.

14. The non-transitory computer-readable medium of claim 8 , wherein the operations further comprise:
generating a second mapping, by the first host machine, based on information associated with each host machine in the cluster and a number of iSCSI targets without affinity owned by the each host machine in the cluster.

15. A first host machine in a cluster of host machines, comprising:
a processor; and
a non-transitory computer-readable storage medium that includes a set of instructions which, in response to execution by the processor, cause the processor to perform operations of object-based load balancing in a distributed storage system of a virtualization system supported by the cluster of host machine

Assignees

• Vmware Inc

Inventors

• Zhou Haitao
• Yang Yang
• Guo Zhaohui
• Zhao Jian
• Huang Zhou
• Feng Jin

Classifications

• G06F13/4221 Primary

  being an input/output bus, e.g. ISA bus, EISA bus, PCI bus, SCSI bus · CPC title

• G06F3/0665

  at area level, e.g. provisioning of virtual or logical volumes · CPC title

• G06F3/0688

  Non-volatile semiconductor memory arrays · CPC title

• G06F9/5083

  Techniques for rebalancing the load in a distributed system · CPC title

• G06F9/45558 Primary

  Hypervisor-specific management and integration aspects · CPC title