Load balancing using a client swapping operation

• US 10,129,331 B2
• Filed: 08/13/2015
• Issued: 11/13/2018
• Est. Priority Date: 06/25/2015
• Status: Active Grant

• Current Assignee: Vmware LLC (Broadcom, Inc.)
• Original Assignee: VMware, Inc. (Broadcom, Inc.)
• Inventors: Jagadeeshwaran, Adarsh, Inabattini, Sai
• Primary Examiner(s): Ndiaye, Cheikh
• Application Number:
  US14/825,193
• Publication Number:
  US 20160381125A1
• Time in Patent Office: 1,188 Days
• CPC Class Codes: H04L 67/1001 for accessing one among a plurality of replicated servers
• CPC Class Codes: H04L 67/1008 based on parameters of servers, e.g. available memory or workload monitoring of computer activity G06F11/30
• CPC Class Codes: H04L 67/1095 Replication or mirroring of data, e.g. scheduling or transport for data synchronisation between network nodes
• CPC Class Codes: H04L 67/1097 for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]

Abstract

A management server and method for load balancing a cluster of host computers analyzes load metrics of clients naming on the host computers in the cluster to select a first client that can be migrated from a first host computer in the cluster to a second host computer in the cluster to improve load balance for the cluster and a second client running on the second host computer that can be swapped with the first client running on the first host computer for a client swapping operation. The client swapping operation involves simultaneously migrating the first client from the first host computer to the second host computer and migrating the second client from the second host computer to the first host computer.

Claims

1. A method for load balancing a cluster of host computers, the method comprising:

analyzing load metrics of clients running on the host computers in the cluster;

selecting a first client that can be migrated from a first host computer in the cluster to a second host computer in the cluster to improve load balance for the cluster based on the analyzing of the load metrics of the clients;

selecting a second client running on the second host computer that can be swapped with the first client running on the first host computer; and

performing a client swapping operation that simultaneously migrate the first client from the first host computer to the second host computer and migrate the second client from the second host computer to the first host computer,

wherein the client swapping operation is an atomic operation performed in an indivisible manner such that both the first and second clients being migrated are returned to their source host computers if the client swapping operation is not fully successful.

2. The method of claim 1, wherein the selecting the first client includes determining that migrating the first client from the first host computer to the second host computer would improve load balance with respect to one resource usage metric but would reduce load balance with respect to another resource usage metric.

3. The method of claim 2, wherein the selecting the second client includes determining that migrating the second client from the second host computer to the first host computer would improve the load balance with respect to the another resource usage metric when the first client is migrated from the first host computer to the second host computer.

4. The method of claim 1, wherein the second host computer is supporting a maximum number of clients allowed for the second host computer when the load metrics of the clients are analyzed.

5. The method of claim 4, wherein the first host computer is supporting a maximum number of clients allowed for the first host computer when the load metrics of the clients are analyzed.

6. The method of claim 4, wherein the maximum number of clients allowed for the second host computer is defined by a setting for the second host computer.

7. The method of claim 1, wherein the performing of the client swapping operation comprises:

creating a first shadow client for the first client on the second host computer;

creating a second shadow client for the second client on the first host computer;

copying memory data of the first client from the first host computer to the second host computer;

copying memory data of the second client from the second host computer to the first host computer;

after copying of the memory data for both first and second clients has been completed, stunning the first and second clients and resuming the first and second clients at their destination host computers using the first and second shadow clients, respectively.

8. The method of claim 7, wherein the copying the memory data of the first client includes executing a plurality of iterations of a pre-copy process, wherein a first iteration includes copying each memory page of the first client from the first host computer to the second host computer and wherein a subsequent iteration includes copying each memory page of the first client that has been changed from a previous iteration from the first host computer to the second host computer.

9. A non-transitory computer-readable storage medium containing program instructions for a method for load balancing a cluster of host computers, wherein execution of the program instructions by one or more processors of a computer system causes the one or more processors to perform steps comprising:

analyzing load metrics of clients running on the host computers in the cluster;

selecting a first client that can be migrated from a first host computer in the cluster to a second host computer in the cluster to improve load balance for the cluster based on the analyzing of the load metrics of the clients;

selecting a second client running on the second host computer that can be swapped with the first client running on the first host computer; and

performing a client swapping operation that simultaneously migrate the first client from the first host computer to the second host computer and migrate the second client from the second host computer to the first host computer,

wherein the client swapping operation is an atomic operation performed in an indivisible manner such that both the first and second clients being migrated are returned to their source host computers if the client swapping operation is not fully successful.

10. The computer-readable storage medium of claim 9, wherein the selecting the first client includes determining that migrating the first client from the first host computer to the second host computer would improve load balance with respect to one resource usage metric but would reduce load balance with respect to another resource usage metric.

11. The computer-readable storage medium of claim 10, wherein the selecting the second client includes determining that migrating the second client from the second host computer to the first host computer would improve the load balance with respect to the another resource usage metric when the first client is migrated from the first host computer to the second host computer.

12. The computer-readable storage medium of claim 9, wherein the second host computer is supporting a maximum number of clients allowed for the second host computer when the load metrics of the clients are analyzed.

13. The computer-readable storage medium of claim 12, wherein the first host computer is supporting a maximum number of clients allowed for the first host computer when the load metrics of the clients are analyzed.

14. The computer-readable storage medium of claim 12, wherein the maximum number of clients allowed for the second host computer is defined by a setting for the second host computer.

15. The computer-readable storage medium of claim 9, wherein the performing of the client swapping operation comprises:

creating a first shadow client for the first client on the second host computer;

creating a second shadow client for the second client on the first host computer;

copying memory data of the first client from the first host computer to the second host computer;

copying memory data of the second client from the second host computer to the first host computer;

after copying of the memory data for both first and second clients has been completed, stunning the first and second clients and resuming the first and second clients at their destination host computers using the first and second shadow clients, respectively.

16. The computer-readable storage medium of claim 15, wherein the copying the memory data of the first client includes executing a plurality of iterations of a pre-copy process, wherein a first iteration includes copying each memory page of the first client from the first host computer to the second host computer and wherein a subsequent iteration includes copying each memory page of the first client that has been changed from a previous iteration from the first host computer to the second host computer.

17. A management server in a distributed computer system with a cluster of host computers, the management server comprising:

memory; and

a processor configured to;

analyze load metrics of clients running on the host computers in the cluster;

select a first client that can be migrated from a first host computer in the cluster to a second host computer in the cluster to improve load balance for the cluster based on the analyzing of the load metrics of the clients;

select a second client running on the second host computer that can be swapped with the first client running on the first host computer; and

perform a client swapping operation that simultaneously migrate the first client from the first host computer to the second host computer and migrate the second client from the second host computer to the first host compute,

wherein the client swapping operation performed by the client migration module is an atomic operation performed in an indivisible manner such that both the first and second clients being migrated are returned to their source host computers if the client swapping operation is not fully successful.

18. The management server of claim 17, wherein the processor is further configured to determine that migrating the first client from the first host computer to the second host computer would improve load balance with respect to one resource usage metric but would reduce load balance with respect to another resource usage metric.

19. The management server of claim 18, wherein the processor is further configured to determine that migrating the second client from the second host computer to the first host computer would improve the load balance with respect to the another resource usage metric when the first client is migrated from the first host computer to the second host computer.

20. The management server of claim 17, wherein at least one of the first and second host computers is supporting a maximum number of clients allowed for that host computer when the load metrics of the clients are analyzed.

21. The management server of claim 17, wherein the processor is further configured to:

create a first shadow client for the first client on the second host computer;

create a second shadow client for the second client on the first host computer;

copy memory data of the first client from the first host computer to the second host computer;

copy memory data of the second client from the second host computer to the first host computer; and

after the memory data for both first and second clients has been copied, stun the first and second clients and resume the first and second clients at their destination host computers using the first and second shadow clients, respectively.