PROXY QUEUE PAIR FOR OFFLOADING

• US 20140181823A1
• Filed: 12/20/2012
• Published: 06/26/2014
• Est. Priority Date: 12/20/2012
• Status: Active Grant

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle International Corporation (Oracle Corporation)
• Inventors: Manula, Brian Edward, Bugge, Haakon Ording, Sandven, Magne Vigulf
• Granted Patent: US 9,069,633 B2
• Time in Patent Office: Days
• US Class Current: 718/102
• CPC Class Codes: G06F 2209/509 Offload
• CPC Class Codes: G06F 2209/548 Queue
• CPC Class Codes: G06F 9/546 Message passing systems or structures, e.g. queues

Abstract

A method for offloading includes a host channel adapter (HCA) receiving a first work request identifying a queue pair (QP), making a first determination that the QP is a proxy QP, and offloading the first work request to a proxy central processing unit (CPU) based on the first determination and based on the first work request satisfying a filter criterion. The HCA further receives a second work request identifying the QP, processes the second work request without offloading based on the QP being a proxy QP and based on the first work request failing to satisfy the filter criterion. The HCA redirects a first completion for the first work request and a second completion for the second work request to the proxy CPU based on the first determination. The proxy CPU processes the first completion and the second completion in order.

Claims

1. A method for offloading work requests of a queue pair (QP), comprising:

receiving, by the HCA, a first work request identifying the QP, wherein the QP is associated with an application executing on a host connected to the HCA;

making a first determination that the QP is a proxy QP;

offloading, by the HCA, the first work request to a proxy central processing unit (CPU) based on the first determination and based on the first work request satisfying a filter criterion;

receiving, by the HCA, a second work request identifying the QP;

processing, by the HCA, the second work request without offloading based on the QP being a proxy QP and based on the first work request failing to satisfy the filter criterion;

redirecting a first completion for the first work request and a second completion for the second work request to the proxy CPU based on the first determination; and

processing, by the proxy CPU, the first completion and the second completion in order.

2. The method of claim 1, wherein the first work request is received from a network and is directed to the application, and wherein offloading the first work request comprises:

redirecting the first work request to the proxy CPU; and

performing, by the proxy CPU, operations of the application according to the first work request.

3. The method of claim 2, wherein processing the first completion and the second completion by the proxy CPU comprises:

transmitting, for the first completion, a no-op instruction to a completion queue (CQ) in host memory located on the host, wherein the CQ is associated with the application; and

transmitting, for the second completion, the second completion to the CQ.

4. The method of claim 2, wherein the operations of the application comprise obtaining, for a remote requestor, a lock on a portion of a distributed database, wherein the portion is located on the host.

5. The method of claim 1, wherein the filter criterion comprises a size of the work request.

6. The method of claim 1, wherein the first work request is received from the application, and wherein offloading the first work request comprises:

redirecting the first work request to the proxy CPU; and

performing, on the first work request, operations of an operating system executing on the host.

7. The method of claim 6, wherein processing the first completion and the second completion by the proxy CPU comprises transmitting the first completion and the second completion in order to a CQ in host memory located on the host, wherein the CQ is associated with the application.

8. The method of claim 1, further comprising:

making a second determination that a CQ associated with the application is a proxy CQ,

wherein redirecting the first completion and the second completion to the proxy CPU is further based on the second determination.

9. A host channel adapter (HCA) for offloading, comprising:

a proxy central processing unit;

a receive module for;

receiving a first work request identifying a first queue pair (QP), wherein the first QP is associated with an application executing on a host connected to the HCA;

making a first determination that the first QP is a proxy QP;

redirecting the first work request to the proxy CPU on the HCA based on the first determination and based on the first work request satisfying a filter criterion;

receiving a second work request identifying the first QP; and

processing the second work request without redirecting to the proxy CPU based on the first determination and based on the first work request failing to satisfy the filter criterion;

a completion module for;

generating a first completion for the first work request; and

generating a second completion for the second work request; and

a descriptor fetch module for;

redirecting the first completion and the second completion to the proxy CPU based on the first determination,

wherein the proxy CPU processes the first work request, the first completion, and the second completion.

10. The HCA of claim 9, wherein the receive module comprises:

a header validation module for;

obtaining a QP state for the first QP; and

making the first determination based on a proxy bit being set in the QP state; and

a redirection module for redirecting work requests to the proxy CPU.

11. The HCA of claim 9, wherein the descriptor fetch module comprises:

a header validation module for;

obtaining a CQ descriptor for a CQ corresponding to the first work request and the second work request; and

making a second determination that the CQ is a proxy CQ based on a proxy bit being set in the CQ descriptor; and

a redirection module for redirecting completions to the proxy CPU.

12. The HCA of claim 9, wherein the first work request is received from a network and is directed to the application, and wherein the proxy CPU performs operations of the application according to the first work request.

13. The HCA of claim 12, wherein processing the first completion and the second completion by the proxy CPU comprises:

transmitting, for the first completion, a no-op instruction to a CQ in host memory located on the host, wherein the CQ is associated with the application; and

transmitting, for the second completion, the second completion to the CQ.

14. The HCA of claim 12, wherein the operations of the application comprises obtaining, for a remote requestor, a lock on a portion of a distributed database, wherein the portion is located on the host.

15. The HCA of claim 12, wherein the proxy CPU is further configured to:

make a second determination that processing of the first work request by the proxy CPU fails; and

transmit the first work request to the application based on the second determination.

16. The HCA of claim 12, wherein the receive module is further configured to:

receive a third work request identifying a second QP;

make a second determination that the second QP is not a proxy QP; and

process the third work request without redirecting to the proxy CPU based on the second determination.

17. The HCA of claim 16, wherein the descriptor fetch module is further configured to:

obtain a third completion for the third work request;

obtain a CQ descriptor for a CQ associated with the third work request; and

generate an error when the CQ descriptor for the third work request indicates that the CQ is not a proxy CQ and when the QP is a proxy QP.

18. A system comprising:

a host comprising an application; and

a host channel adapter (HCA), connected to the host, for;

receiving a first work request identifying a queue pair (QP), wherein the QP is associated with the application;

making a first determination that the QP is a proxy QP;

offloading the first work request to a proxy central processing unit (CPU) based on the first determination and based on the first work request satisfying a filter criterion;

receiving a second work request identifying the QP;

processing the second work request without offloading based on the QP being a proxy QP and based on the first work request failing to satisfy the filter criterion;

redirecting a first completion for the first work request and a second completion for the second work request to the proxy CPU based on the first determination; and

processing, by the proxy CPU, the first completion and the second completion in order.

19. The system of claim 18, wherein the first work request is received from a network and is directed to the application, and wherein offloading the first work request comprises:

redirecting the first work request to the proxy CPU; and

performing operations of the application by the proxy CPU according to the first work request.

20. The system of claim 19, wherein processing the first completion and the second completion by the proxy CPU comprises:

transmitting, for the first completion, a no-op instruction to a CQ in host memory located on the host, wherein the CQ is associated with the application; and

transmitting, for the second completion, the second completion to the CQ.