Granted Patent B1

US 8,867,537 · App. 13/610,095 · Granted Oct 21, 2014

Abstract

Claims (48)

1. A method for performing a read operation, the method comprising:

a first computing system registering a first memory buffer with a first network device, and a second computing system registering a second memory buffer with a second network device;

providing a first unique identifier for the first memory buffer, and a second unique identifier for the second memory buffer;

the second computing system advertising the second memory buffer to the first computing system;

the first computing system generating a read request to read information from a memory location and an untagged buffer PDU (Protocol Data Unit);

the first computing system sending the read request and the untagged buffer PDU to the second computing system;

transferring data from the second buffer to a tagged buffer PDU;

sending the tagged buffer PDU to the first computing system;

moving a payload of the tagged buffer PDU directly to the first memory buffer;

notifying the first computing system that the payload of the tagged buffer PDU has been moved to the first memory buffer, and creating an untagged buffer acknowledgement PDU; and

sending the untagged buffer acknowledgement PDU to the second computing system.

2. The method of claim 1 , wherein the first network device is a remote direct memory access (RDMA) enabled network interface card (RNIC) of the first computing system, and the first RNIC provides the first unique identifier, and the second network device is a second RNIC of the second computing system, where the second RNIC provides the second unique identifier.

3. The method of claim 1 , wherein an RNIC of the first computing system creates the untagged buffer PDU.

4. The method of claim 1 , wherein the untagged buffer PDU includes the first and second unique identifiers.

5. The method of claim 1 , wherein at least one check is performed prior to transferring the data from the second buffer to the tagged buffer PDU.

6. The method of claim 5 , wherein the at least one check includes at least one of CRC (Cyclic Redundancy Check), checking the first and/or second unique identifiers, and checking an offset.

7. The method of claim 1 , wherein the first network device of the first computing system notifies the first computing system that the payload of the tagged buffer PDU has been moved to the first memory buffer.

8. The method of claim 1 , wherein the tagged buffer PDU sent to the first computing system is sent to the first network device of the first computing system.

9. The method of claim 1 , further comprising:

adding the sent read request PDU to a send timer list;

queuing a read response in the send timer list;

removing the read request from the send timer list; and

after the untagged buffer acknowledgement PDU is received, removing the read request from the send timer list;

wherein the second network device of the second computing system removes the read response that was added to the send timer list.

10. A system for performing a read operation, the system comprising:

a first computing system having a first memory buffer with a first unique identifier, and a second computing system having a second memory buffer with a second unique identifier;

wherein the system is configured to

advertise the second memory buffer to the first computing system;

generate a read request to read information from a memory location and an untagged buffer PDU (Protocol Data Unit) in the first computing system, and send the read request and the untagged buffer PDU to the second computing system;

transfer data from the second buffer to a tagged buffer PDU;

send the tagged buffer PDU to the first computing system;

move a payload of the tagged buffer PDU directly to the first memory buffer;

notify the first computing system that the payload of the tagged buffer PDU has been moved to the first memory buffer, and create an untagged buffer acknowledgement PDU; and

send the untagged buffer acknowledgement PDU to the second computing system.

11. The system of claim 10 , further comprising a first network device of the first computing system and a second network device of the second computing system.

12. The system of claim 11 , wherein the first network device is a remote direct memory access (RDMA) enabled network interface card (RNIC), and the first RNIC provides the first unique identifier, and the second network device is a second RNIC, and the second RNIC provides the second unique identifier.

13. The system of claim 11 , wherein an RNIC of the first computing system creates the untagged buffer PDU.

14. The system of claim 10 , wherein the untagged buffer PDU includes the first and second unique identifiers.

15. The system of claim 10 , wherein at least one check is performed prior to transferring the data from the second memory buffer to the tagged buffer PDU.

16. The system of claim 15 , wherein the at least one check includes at least one of CRC (Cyclic Redundancy Check), checking the first and/or second unique identifiers, and checking an offset.

17. The system of claim 11 , wherein the first network device notifies the first computing system that the payload of the tagged buffer PDU has been moved to the first memory buffer.

18. The system of claim 11 , wherein the tagged buffer PDU sent to the first computing system is sent to the first network device.

19. The system of claim 10 , further comprising:

adding the sent read request PDU to a send timer list;

queuing a read response in the send timer list;

removing the read request from the send timer list; and

after the untagged buffer acknowledgement PDU is received, removing the read request from the send timer list;

wherein the second network device of the second computing system removes the read response that was added to the send timer list.

Assignments (7)