TCP/IP offload engine virtualization system and methods

• US 8,028,071 B1
• Filed: 02/14/2007
• Issued: 09/27/2011
• Est. Priority Date: 02/15/2006
• Status: Active Grant

• Current Assignee: Vmware LLC (Broadcom, Inc.)
• Original Assignee: VMware, Inc. (Broadcom, Inc.)
• Inventors: Arunachalam, Ramu, Lambeth, Walter Andrew, Mahalingam, Mallik
• Primary Examiner(s): Backer; Firmin
• Assistant Examiner(s): Chambers; Michael A
• Application Number:
  US11/705,842
• Time in Patent Office: 1,686 Days
• Field of Search: 709/1, 709/239, 709/250, 709/228, 709/227, 718/1, 370/254, 370/389
• US Class Current: 709/227
• CPC Class Codes: G06F 2009/45595 Network integration; Enabling network access in virtual machine instances
• CPC Class Codes: G06F 9/45558 Hypervisor-specific management and integration aspects
• CPC Class Codes: H04L 69/32 Architecture of open systems interconnection [OSI] 7-layer type protocol stacks, e.g. the interfaces between the data link level and the physical level

Abstract

A virtual computer system enabling dynamic, aggregated use of multiple TCP/IP offload engines (TOEs) by the set of guest computer systems hosted on the virtual computer system. Each of the guest computer systems includes an offload selection switch and the associated virtual machine monitor includes a first virtual context component. Second virtual context components are associated with a set of TCP/IP stacks and TOEs and interoperate with the first virtual context components to establish a virtual routing of network connections between the offload selection switches and the TOEs. The virtual context mapping retains the initially requested network connection information as well as the resolved virtual network connection established, thereby allowing the initial network connection request to be internally reapplied as required to accommodate dynamic changes in the network protocol parameters of the TOEs.

Claims

1. A virtual computer system implementing architectural support for the use of one or more TCP/IP offload engines (TOE) by the guest computer systems hosted on the virtual computer system, each of the guest computer systems including a guest operating system and each guest operating system including a network stack, said virtual computer system comprising:

a) a host platform including a central processing unit, system memory, a plurality of network interface controllers (NICs), and a plurality of TOEs, wherein said plurality of NICs and said plurality of TOEs is coupleable to one or more physical communications network links;

b) a plurality of guest computer system components wherein each guest computer system component includes an offload selection switch provided in the corresponding guest operating system and a first virtual context component, wherein each offload selection switch selectively establishes network connections for applications in each corresponding guest computer system with the corresponding first virtual context component or the network stack; and

c) a virtual computer system kernel including a plurality of second virtual context components having a first defined correspondence with said first virtual context components, a network hardware abstraction layer, and a plurality of TOE drivers having a second defined correspondence with said plurality of TOEs, each second virtual context component defined to select and establish a route to a TOE driver for servicing a request for a communications session, wherein a virtual TOE device is implemented as a pair of one of the first virtual context components and one of the second virtual context components, wherein said first and second virtual context components interoperate to maintain virtual associations of the communications sessions established through said first and second virtual context components, wherein said virtual associations include requested and resolved associations and where said requested and resolved associations include the ports and IP addresses that define the communications session.

2. The virtual computer system of claim 1 further including a persistent state store coupled to said virtual computer system kernel, wherein said persistent state store provides for the storage of a predetermined set of virtual associations.

3. The virtual computer system of claim 1 wherein said first and second virtual context components interoperate to alter existent virtual associations in response to a changed network protocol parameter of the corresponding communications session, whereby established network sessions can be maintained transparently to participating application programs.

4. The virtual computer system of claim 3 wherein said first virtual context components support deferred binding of network session connections relative to application programs and recording of the option configurations requested by application programs in establishing network session connections, said first virtual context components interoperating with said second virtual context components to resolve and establish network session connections, said first virtual context components being operative to replay the option configuration requests for a predetermined network communication session once the network session connection is resolved.

5. The virtual computer system of claim 4 wherein first and second virtual context components interoperate to define a bound socket connection between a socket supported by said first virtual context component and a socket supported by at least one TOE of said plurality of TOEs.

6. The virtual computer system of claim 5 wherein said second virtual context components provide for the selection and establishment of a socket connection with a dynamically selected TOE of said plurality of TOEs, where in the dynamic selection is dependent on the current network parameters of each TOE of said plurality of TOEs.

7. The virtual computer system of claim 6 wherein said second virtual context components are responsive to alterations in the current network parameters of each of said plurality of TOEs, wherein said second virtual context components selectively redefine bound socket connections between sockets supported by said first virtual context component and sockets supported by a dynamically selected TOE of said plurality of TOEs.

8. The virtual computer system of claim 7 further including a state store coupled to said virtual computer system kernel, wherein said state store provides for the storage of a predetermined set of virtual associations, and wherein said first virtual context component selectively provides for the redetermination of said bound socket connection between said socket supported by said first virtual context component and another socket supported by at least one TOE of said plurality of TOEs based on said requested associations stored by said state store.

9. The virtual computer system of claim 8 wherein said state store includes a persistent state store.

10. A method of enabling shared use of one or more TCP/IP offload engines (TOE) by a plurality of guest computer systems hosted on a virtual computer system, each of the guest computer systems including a guest operating system and each guest operating system including a network stack, said method comprising the steps of:

providing an offload selection switch and a first component layer in each of the plurality of guest computer systems hosted on the virtual computer system, wherein each offload selection switch selectively establishes network connections for applications in each guest computer system with the corresponding first component layer or the network stack;

providing a second component layer on said virtual computer system, wherein an instance of said second component layer is associated with a TOE driver linked to a corresponding TOE coupled to said virtual computer system; and

wherein said first component layers receive information sets defining requested sessions;

wherein, for a given requested session, the corresponding first component layer communicates with a selected one of said second component layers to determine for said given requested session a virtual session; and

wherein said selected second component layer communicates through an associated TOE driver to establish a network session conforming to said virtual session.

11. The method of claim 10 further comprising the steps of:

a) storing said information sets defining requests sessions; and

b) selectively using said information sets internal to said first and second component layers to reselect one of said second component layers as said selected one of said second component layers and redetermine said virtual session for said given requested session, wherein said selected second component layer communicates through said associated TOE driver to establish a network session conforming to said virtual session.

12. The method of claim 11 further comprising the steps of:

a) receiving, by said first component layers, notice of changes to predetermined network parameters maintained by said TCP/IP offload engines; and

b) initiating, selectively in response to a notice of a predetermined change in said predetermined network parameters, said step of selectively using.

13. The method of claim 12 further comprising the step of persistently storing the information sets corresponding to said requested sessions.

14. The method of claim 12 wherein said information sets include the applicable port, IP, and MAC addresses defining the requested network session.

15. The method of claim 14 wherein said information sets further include control parameters used to define the operation of the requested network session.

16. The method of claim 15 wherein said step of selectively using includes re-application of the control parameters in the establishment of the requested network session.

17. A virtual computer system enabling aggregated use of one or more TCP/IP offload engines (TOE) by a plurality of guest computer systems hosted on the virtual computer system, each of the guest computer systems including a guest operating system and each guest operating system including a network stack, said virtual computer system comprising:

a) a host platform including a central processing unit, system memory, a plurality of network interface controllers (NICs), and a plurality of TOEs, wherein said plurality of NICs and said plurality of TOEs is coupleable to one or more physical communications network links;

b) a plurality of guest computer system components wherein each guest computer system component includes an offload selection switch provided in the corresponding guest operating system;

c) a plurality of virtual machine monitors coupled respectively to said plurality of guest computer system components, each of said plurality of virtual machine monitors including a first virtual context component coupled to a respective one of said offload selection switches to selectively establish network connections for applications in the corresponding guest computer system with the corresponding first virtual context component or the network stack, wherein said first virtual context components maintain a mapping between application network request contexts specific to corresponding ones of said plurality of guest computer system components and virtual network request contexts corresponding to said plurality of TOEs; and

d) a virtual computer system kernel including a plurality of network stacks and TOE engine drivers coupled respectively to said plurality of TOEs, a plurality of second virtual context components having a defined correspondence with said plurality of TOEs, wherein said plurality of second virtual context components is coupled to and interoperative with said plurality of virtual machine monitors to determine said mapping.

18. The virtual computer system of claim 17 wherein said mapping includes, for a given network session request, a given request defined port number, IP address specification, and MAC address, as received by a corresponding one of said offload selection switches, and a resolved port number, IP address specification, and MAC address, as determined by a corresponding one of said first virtual context components and a predetermined one of said second virtual context components.

19. The virtual computer system of claim 18 wherein said first virtual context components are responsive to predetermined network protocol parameters maintained by said plurality of network stacks and wherein said first virtual context components select said second virtual context components for interoperation in defining said mapping based on said predetermined network protocol parameters.

20. The virtual computer system of claim 19 wherein said first virtual context components are dynamically responsive to changes in said predetermined network protocol parameters to reselect, for said given network session request, a corresponding one of said second virtual context components and redetermine said resolved port number, IP address specification and MAC address.

21. A virtual computer system implementing architectural support for the use of one or more TCP/IP offload engines (TOE) by the guest computer systems hosted on the virtual computer system, each of the guest computer systems including a guest operating system and each guest operating system including a network stack, said virtual computer system comprising:

a) a host platform including a plurality of network interface controllers (NICs) and a plurality of TOE engines, wherein said plurality of NICs and said plurality of TOEs is coupleable to one or more physical communications network links;

b) a plurality of guest computer system components wherein each guest computer system component includes an offload selection switch provided in the corresponding guest operating system and a first virtual context component, wherein each offload selection switch selectively establishes network connections for applications in each corresponding guest computer system with the corresponding first virtual context component or the network stack; and

c) a virtual computer system kernel including a plurality of second virtual context components and a plurality of TOE drivers in communication with said plurality of TOEs, each second virtual context component defined to select and establish a route to a TOE driver for servicing a request for a communications session, wherein a virtual TOE device is implemented as a pair of one of the first virtual context components and one of the second virtual context components.