Zero-copy network and file offload for web and application servers

• US 8,713,180 B2
• Filed: 03/22/2006
• Issued: 04/29/2014
• Est. Priority Date: 06/22/2005
• Status: Active Grant

• Current Assignee: Cisco Technology, Inc. (Cisco Systems, Inc.)
• Original Assignee: Cisco Technology, Inc. (Cisco Systems, Inc.)
• Inventors: Bakke, Mark A., Kuik, Timothy J., Thompson, David P.
• Primary Examiner(s): MAI, KEVIN S
• Application Number:
  US11/386,487
• Publication Number:
  US 20060294234A1
• Time in Patent Office: 2,960 Days
• Field of Search: 709/203, 709/213, 709223-224, 709/250, 370/401, 370/469
• US Class Current: 709/226
• CPC Class Codes: H04L 69/16 Implementation or adaptation of Internet protocol [IP], of transmission control protocol [TCP] or of user datagram protocol [UDP]
• CPC Class Codes: H04L 69/161 Implementation details of TCP/IP or UDP/IP stack architecture; Specification of modified or new header fields
• CPC Class Codes: H04L 69/162 involving adaptations of sockets based mechanisms secure socket layer H04L63/168
• 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
• CPC Class Codes: H04L 69/326 in the transport layer [OSI layer 4] H04L69/16 takes precedence

Abstract

Methods and apparatus for transferring data from an application server are provided. By offloading network and file system stacks to a common stack accessible by multiple operating systems in a virtual computing system, embodiments of the present invention may achieve data transfer support for web and application servers without the data needing to be copied to or reside in the address space of the server operating systems.

Claims

1. A computer-implemented method to transfer data via a buffer cache and in response to a command issued from a first application server, the command issued responsive to a request from a first client, the method comprising:

creating an offload stack running on a multiprocessing host having the buffer cache and on which a plurality of operating systems is running, wherein the plurality of operating systems includes a first operating system executing the first application server, wherein the offload stack is configured to provide a set of resources to one or more applications executing on the plurality of operating systems, the set of resources including network, file, and block I/O resources;

providing an interface to share the set of resources of the offload stack among the plurality of operating systems running on the multiprocessing host; and

providing zero-copy data transfer support for the first application server by processing the command issued from the first application server to the offload stack, in order to transfer data targeted by the command, to the first client, via the buffer cache of the multiprocessing host and not via any address space of the first operating system on which the first application server is running, wherein the command specifies a source of the data to transfer, a destination to transfer the data to, and an offset to start reading the data from.

2. The method of claim 1, wherein the command includes at least a source argument to specify the source of the data to be transferred and a destination argument to specify the destination to which the data is to be transferred.

3. The method of claim 1, wherein the interface to share the set of resources of the offload stack among the plurality of operating systems running on the multiprocessing host comprises driver code executable by the plurality of operating systems.

4. The method of claim 1, wherein the offload stack is created on a data path processor separate from any computer processor on which any of the plurality of operating systems are running.

15. The method of claim 1, wherein the interface to share the set of resources of the offload stack between the multiple operating systems running on the multiprocessing host comprises driver code executable by the plurality of operating systems.

16. The method of claim 15, wherein the offload stack is created on a data path processing element separate from any processing element on which any of the multiple operating systems are running, wherein each processing element is selected from a computer processor and a processing core.

17. The method of claim 16, wherein the command issued from the first application server is issued responsive to the first application server receiving the request from the first client, wherein the first client comprises a remote client of the first application server.

18. The method of claim 17, wherein the method further comprises:

providing zero-copy data transfer support for a second application server by processing a command issued from the second application server to the offload stack, in order to transfer data targeted by the command, to a second client, via the buffer cache of the multiprocessing host and not via any address space of a second operating system on which the second application server is running.

19. The method of claim 18, wherein the second operating system is different from the first operating system, wherein the second application is different from the first application, wherein the second client is different from the first client, wherein the command issued from the second application server is issued responsive to the second application server receiving a request from the second client, wherein the second client comprises a remote client of the second application.

20. The method of claim 19, wherein the first operating system and the second operating system are of different operating system types, wherein the first application and the second application are of different application types, wherein the offload stack does not occupy any address space of the first operating system.

21. The method of claim 20, wherein the data is transferred without copying the data from the buffer cache to any kernel address space of the first operating system, without copying the data from any kernel address space of the first operating system to any user address space of the first operating system, and without copying the data from any user address space of the first operating system to any socket buffer of the first operating system.

22. The method of claim 21, wherein the multiprocessing host consists of a single host system, wherein the multiprocessing host includes a plurality of processing elements, wherein the offload stack runs solely on a first processing element of the plurality of processing elements, wherein the first operating system runs solely on a second processing element of the plurality of processors elements, wherein the first processing element is different from the second processing element.

23. The method of claim 22, wherein the data targeted by the command is transferred to the first client using a socket connection via a socket buffer of the multi-core host, wherein the socket buffer is different from the buffer cache, wherein the socket buffer comprises a TCP socket buffer.

24. The method of claim 23, wherein the buffer cache does not occupy any address space of the first operating system, wherein the buffer cache does not occupy any address space of the second operating system, wherein the socket buffer does not occupy any address space of the first operating system, wherein the socket buffer does not occupy any address space of the second operating system.

5. A host system to transfer data via a buffer cache and in response to a command issued from a first application server, the command issued responsive to a request from a first client, the host system comprising:

one or more processing cores;

the buffer cache;

multiple operating systems running on at least one of the one or more processing cores;

a first application server running on a first operating system of the multiple operating systems;

an offload stack running on at least one of the one or more processing cores, the offload stack providing zero-copy data transfer support for the first application server by processing the command issued from the first application server, in order to transfer data targeted by the command, to the first client, via the buffer cache of the host system and not via any address space of the first operating system on which the first application server is running, wherein the offload stack is configured to provide a set of resources to one or more applications executing on the plurality of operating systems, the set of resources including network, file, and block I/O resources, wherein the command specifies a source of the data to transfer, a destination to transfer the data to, and an offset to start reading the data from; and

an interface to share the set of resources of the offload stack between the multiple operating systems.

6. The host system of claim 5, wherein the host system includes a plurality of processing cores including the one or more processing cores, wherein the first application server and the offload stack are running on different processing cores.

7. The host system of claim 5, wherein the command includes at least a source argument to specify the source of the data to be transferred and a destination argument to specify the destination to which the data is to be transferred.

8. The host system of claim 5, wherein the offload stack runs on a data path processing core separate from any processing core on which any of the multiple operating systems are running.

9. A system to transfer data via a buffer cache and in response to a command issued from a first application server, the command issued responsive to a request from a first client, the system comprising:

one or more computer processors;

a buffer cache;

a memory containing a program which, when executed by the one or more computer processors, performs an operation comprising;

creating an offload stack running on the system, on which system a plurality of operating systems is running, the plurality of operating systems including a first operating system executing the first application server, wherein the offload stack is configured to provide a set of resources to one or more applications executing on the plurality of operating systems, the set of resources including network, file, and block I/O resources;

providing an interface to share the set of resources of the offload stack among the plurality of operating systems running on the system; and

providing zero-copy data transfer support for the first application server by processing the command issued from the first application server to the offload stack, in order to transfer data targeted by the command, to the first client, via the buffer cache of the system and not via any address space of the first operating system on which the first application server is running, wherein the command specifies a source of the data to transfer, a destination to transfer the data to, and an offset to start reading the data from.

10. The system of claim 9, wherein the command includes at least a source argument to specify the source of the data to be transferred and a destination argument to specify the destination to which the data is to be transferred.

11. The system of claim 10, wherein the interface to share the set of resources of the offload stack among the plurality of operating systems running on the system comprises driver code executable by the plurality of operating systems.

12. The system of claim 11, wherein the offload stack is created on a data path processor separate from any computer processor on which any of the plurality of operating systems are running.

13. A non-transitory computer-readable medium containing program code which, when executed, performs an operation to transfer data via a buffer cache and in response to a command issued from a first application server, the command issued responsive to a request from a first client, the operation comprising:

creating an offload stack running on a multiprocessing host having the buffer cache and on which a plurality of operating systems is running, wherein the plurality of operating systems includes a first operating system executing the first application server, wherein the offload stack is configured to provide a set of resources to one or more applications executing on the plurality of operating systems, the set of resources including network, file, and block I/O resources;

providing an interface to share the set of resources of the offload stack among the plurality of operating systems running on the multiprocessing host; and

providing zero-copy data transfer support for the first application server by processing the command issued from the first application server to the offload stack, in order to transfer data targeted by the command, to the first client, via the buffer cache of the multiprocessing host and not via any address space of the first operating system on which the first application server is running, wherein the command specifies a source of the data to transfer, a destination to transfer the data to, and an offset to start reading the data from.

14. The non-transitory computer-readable medium of claim 13, wherein the offload stack is created on a data path processor separate from any computer processor on which any of the plurality of operating systems are running.