System and method for supporting live migration of virtual machines in an infiniband network

Patent metadata

Field	Value
Publication number	US-9397954-B2
Application number	US-201313837922-A
Country	US
Kind code	B2
Filing date	Mar 15, 2013
Priority date	Mar 26, 2012
Publication date	Jul 19, 2016
Grant date	Jul 19, 2016

Abstract

A system and method can support virtual machine live migration in a network. A plurality of hosts can be interconnected via the network. Each host contains one or more virtual machines. At least one virtual machine operates to perform a live migration from a first host to a second host. The virtual function, which is attached to a queue pair (QP), can be supported by a host channel adapter (HCA) on a host.

First claim

What is claimed is:

1. A system for supporting virtual machine live migration in a network, comprising:
one or more microprocessors;
a plurality of hosts, running on the one or more microprocessors, wherein the plurality of hosts are interconnected via the network;
one or more virtual machines, wherein at least one virtual machine in the one or more virtual machines operates to perform a migration of the at least one virtual machine from a first host to a second host,
wherein said at least one virtual machine is attached with a virtual function that is associated with a queue pair (QP), wherein the QP is supported by a host channel adapter (HCA) at the first host, and wherein one or more attributes of the queue pair (QP) associated with the HCA at the first host are registered in a table; and
wherein during the migration of the at least one virtual machine from the first host to the second host
the virtual function is detached from the virtual machine at the first host;
a queue pair (QP) is newly created for use by the second host, using the one or more attributes registered in the table, and is associated with a HCA at the second host:
the virtual machine is migrated to the second host; and
a new virtual function is attached to the migrated virtual machine and the newly created QP at the second host.

2. The system according to claim 1 , further comprising:
at least one virtual machine monitor that manages the one or more virtual machines.

3. The system according to claim 1 , wherein:
said virtual function is provided in an input/output (I/O) virtualization model.

4. The system according to claim 1 , wherein:
said at least one virtual machine includes
a user space, wherein the user space includes one or more user processes, each of which is associated with a process identification number (PID), and
a kernel space, wherein the kernel space includes a translation table as the table and maps a PID to a queue pair number (QPN), and wherein the translation table is referenced to recreate the queue pair (QP) with the one or more attributes on the second host when the virtual function is detached from the at least one virtual machine on the first host and the new virtual function is attached to the migrated virtual machine on the second host.

5. The system according to claim 1 , wherein:
a late-detach migration process is used to reduce network/service down-time during the migration of the at least one virtual machine from the first host to the second host.

6. The system according to claim 1 , wherein:
one or more location-dependent QP attributes can be updated using the table after the at least one virtual machine is migrated from the first host to the second host.

7. The system according to claim 6 , wherein:
each user process in the user space include a mapping table as the table and operates to search and retrieve the updated one or more location-dependent QP attributes for setup of communications of the queue pair (QP) on the second host.

8. The system according to claim 1 , wherein:
QP state information can be migrated along with an image of said at least one virtual machine in order to reallocate physical resources of QPs.

9. The system according to claim 1 , wherein:
a peer QP can halt communication with the QP during the migration of said at least one virtual machine.

10. The system according to claim 9 , wherein:
a connection manager (CM) can reestablish a new connection between the migrated QP and the peer QP after the migration is completed.

11. The system according to claim 1 , wherein
the table is used to maintain application transparency and pin virtual addresses in the user process to physical memory at a new location associated with the second host after the migration.

12. A method for supporting virtual machine live migration in a network, comprising:
providing a plurality of hosts, wherein the plurality of hosts are interconnected via the network;
providing one or more virtual machines, wherein the at least one virtual machine in the one or more virtual machines is attached with a virtual function that is associated with a queue pair (QP) in a first host, wherein the QP is supported by a host channel adapter (HCA) at the first host;
registering one or more attributes of the queue pair (QP) associated with the HCA at the first host in a table; and
performing, via said at least one virtual machine, a migration from the first host to a second host that includes:
detaching the virtual function from the virtual machine at the first host;
using the one or more attributes registered in the table to newly create a queue pair (QP) for use by the second host and is associated with a HCA at the second host;
migrating the virtual machine to the second host; and
attaching a new virtual function to the migrated virtual machine and the newly created QP at the second host.

13. The method according to claim 12 , further comprising:
providing at least one virtual machine monitor that manages the one or more virtual machines.

14. The method according to claim 12 , further comprising:
including a user space in said at least one virtual machine, wherein the user space includes one or more user processes, each of which is associated with a process identification number (PID), and
including a kernel space in said at least one virtual machine, wherein the kernel space includes a translation table as the table and maps a PID to a queue pair number (QPN), and wherein the translation table is referenced to recreate the queue pair (QP) with the attribute on the second host when the virtual function is detached from the at least one virtual machine on the first host and the new virtual function is attached to the migrated virtual machine on the second host.

15. The method according to claim 12 , further comprising:
using a late-detach migration process to reduce network/service down-time during the migration of the at least one virtual machine from the first host to the second host.

16. The method according to claim 12 , further comprising:
updating one or more location-dependent QP attributes using the table after the at least one virtual machine is migrated from the first host to the second host.

17. The method according to claim 16 , further comprising:
using a mapping table as the table and associating the mapping table with each user process in the a user space of the at least one virtual machine, wherein the mapping table operates to search and retrieve the updated one or more location-dependent QP attributes for the setup of communications of the queue pair (QP) on the second host.

18. The method according to claim 12 , further comprising:
migrating QP state information along with an image of said at least one virtual machine in order to reallocate physical resources of QPs.

19. The method according to claim 12 , further comprising:
configuring a peer QP to halt communication with the QP during the migration of said at least one virtual machine.

20. A non-transitory machine readable storage medium having instructions stored thereon that when executed cause a system to perform the steps comprising:
providing a plurality of hosts, wherein the plurality of hosts are inter-connected via the network;
providing one or more virtual machines, wherein the at least one virtual machine in the one or more virtual machines is attached with a virtual function that is associated with a queue pair (QP) in a first host, wherein the QP is supported by a host channel adapter (H

Assignees

• Oracle Int Corp

Inventors

• Guay Wei Lin
• Johnsen Bjørn Dag

Classifications

• G06F9/4856 Primary

  resumption being on a different machine, e.g. task migration, virtual machine migration (G06F9/5088 takes precedence) · CPC title

• H04L45/02

  Topology update or discovery · CPC title

• H04L49/358

  Infiniband Switches · CPC title

• G06F9/45533

  Hypervisors; Virtual machine monitors · CPC title

• G06F2009/4557

  Distribution of virtual machine instances; Migration and load balancing · CPC title