Resource affinity via dynamic reconfiguration for multi-queue network adapters

• US 8,589,941 B2
• Filed: 04/23/2010
• Issued: 11/19/2013
• Est. Priority Date: 04/23/2010
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Cardona, Omar, Cunningham, James B., Ochs, Matthew R., Sharma, Rakesh
• Primary Examiner(s): Sough, H S
• Assistant Examiner(s): Onat, Umut
• Application Number:
  US12/766,282
• Publication Number:
  US 20110265095A1
• Time in Patent Office: 1,306 Days
• Field of Search: None
• US Class Current: 718/105
• CPC Class Codes: G06F 9/505 considering the load
• CPC Class Codes: H04L 41/082 the condition being updates or upgrades of network functionality
• CPC Class Codes: H04L 61/5014 using dynamic host configuration protocol [DHCP] or bootstrap protocol [BOOTP]

Abstract

A mechanism is provided for providing resource affinity for multi-queue network adapters via dynamic reconfiguration. A device driver allocates an initial queue pair within a memory. The device driver determines whether workload of the data processing system has risen above a predetermined high threshold. Responsive to the workload rising above the predetermined high threshold, the device driver allocates and initializes an additional queue pair in the memory. The device driver programs a receive side scaling (RSS) mechanism in a network adapter to allow for dynamic insertion of an additional processing engine associated with the additional queue pair. The device driver enables transmit tuple hashing to the additional queue pair.

Claims

1. A computer program product comprising a computer readable storage medium having a computer readable program stored therein, wherein the computer readable program, when executed on a computing device, causes the computing device to:

allocate an initial queue pair within a memory, wherein each queue pair is a transmit/receive queue pair;

determine whether workload of the computing device has risen above a predetermined high threshold;

responsive to the workload rising above the predetermined high threshold;

allocate and initialize an additional queue pair in the memory;

program a receive side scaling (RSS) mechanism in a network adapter to allow for dynamic insertion of an additional processing engine associated with the additional queue pair; and

enable transmit tuple hashing to the additional queue pair;

wherein the computer readable program further causes the computing device to;

determine whether the workload has fallen below a predetermined low threshold;

responsive to the workload falling below the predetermined low threshold, determine whether there is only one queue pair remaining allocated in the memory; and

responsive to more than one queue pair remaining allocated in the memory;

reprogram the RSS mechanism in the network adapter to allow for deletion of an allocated queue pair;

disable transmit tuple hashing to an identified queue pair;

determine whether the workload to the identified queue pair has quiesced; and

responsive to the workload to the identified queue pair quiescing, remove the identified queue pair from memory, thereby freeing up memory used by the identified queue pair.

2. The computer program product of claim 1, wherein the computer readable program further causes the computing device to:

repeat the computer readable program to allocate and initialize, program, and enable each time the workload of the computing device rises above the predetermined high threshold.

3. The computer program product of claim 1, wherein the computer readable program to program the receive side scaling (RSS) mechanism in the network adapter to allow for dynamic insertion of the additional processing engine associated with the additional queue pair further causes the computing device to:

enable transmit tuple hashing to the initial queue pair.

4. The computer program product of claim 1, wherein the computer readable program to determine whether the workload of the computing device has risen above the predetermined high threshold is performed by the computer readable program that further causes the computing device to monitor the workload of the computing device through data flow and resource availability.

5. The computer program product of claim 4, wherein the computer readable program further causes the computing device to monitor at least one parameter associated with the workload and wherein the at least one parameter is at least one of transmit/receive bytes per second, a number of flow control frames being sent and received by the network adapter, a number of direct memory access (DMA) overruns detected by the network adapter, a number of transmit timeout events detected by a device driver, a number of receive packets processed by the device driver per interrupt, or a number of transmit packets on a software queue.

6. The computer program product of claim 1, wherein the computer readable program further causes the computing device to:

responsive to the workload to the identified queue pair failing to quiesce, wait for the workload to the identified queue pair to quiesce before removing the identified queue pair from memory, thereby freeing up memory used by the identified queue pair.

7. An apparatus, comprising:

a processor; and

a memory coupled to the processor, wherein the memory comprises instructions which, when executed by the processor, cause the processor to;

allocate an initial queue pair within a memory, wherein each queue pair is a transmit/receive queue pair;

determine whether workload of the apparatus has risen above a predetermined high threshold;

responsive to the workload rising above the predetermined high threshold;

allocate and initialize an additional queue pair in the memory;

program a receive side scaling (RSS) mechanism in a network adapter to allow for dynamic insertion of an additional processing engine associated with the additional queue pair; and

enable transmit tuple hashing to the additional queue pair;

wherein the instructions further cause the processor to;

determine whether the workload has fallen below a predetermined low threshold;

responsive to the workload falling below the predetermined low threshold, determine whether there is only one queue pair remaining allocated in the memory; and

responsive to more than one queue pair remaining allocated in the memory;

reprogram the RSS mechanism in the network adapter to allow for deletion of an allocated queue pair;

disable transmit tuple hashing to an identified queue pair;

determine whether the workload to the identified queue pair has quiesced; and

responsive to the workload to the identified queue pair quiescing, remove the identified queue pair from memory, thereby freeing up memory used by the identified queue pair.

8. The apparatus of claim 7, wherein the instructions further cause the processor to:

repeat the computer readable program to allocate and initialize, program, and enable each time the workload of the apparatus rises above the predetermined high threshold.

9. The apparatus of claim 7, wherein the instructions to program the receive side scaling (RSS) mechanism in the network adapter to allow for dynamic insertion of the additional processing engine associated with the additional queue pair further cause the processor to:

enable transmit tuple hashing to the initial queue pair.

10. The apparatus of claim 7, wherein the instructions to determine whether the workload of the apparatus has risen above the predetermined high threshold is performed by the instructions that further cause the processor to monitor the workload of the apparatus through data flow and resource availability.

11. The apparatus of claim 10, wherein the instructions further causes the processor to monitor at least one parameter associated with the workload and wherein the at least one parameter is at least one of transmit/receive bytes per second, a number of flow control frames being sent and received by the network adapter, a number of direct memory access (DMA) overruns detected by the network adapter, a number of transmit timeout events detected by a device driver, a number of receive packets processed by the device driver per interrupt, or a number of transmit packets on a software queue.

12. The apparatus of claim 7, wherein the instructions further cause the processor to:

responsive to the workload to the identified queue pair failing to quiesce, wait for the workload to the identified queue pair to quiesce before removing the identified queue pair from memory, thereby freeing up memory used by the identified queue pair.