Interrupt distribution of a single flow across multiple processors

• US 10,728,167 B2
• Filed: 08/10/2018
• Issued: 07/28/2020
• Est. Priority Date: 08/10/2018
• Status: Active Grant

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle International Corporation (Oracle Corporation)
• Inventors: Singh, Brijesh
• Primary Examiner(s): Divito, Walter J
• Application Number:
  US16/100,309
• Publication Number:
  US 20200053019A1
• Time in Patent Office: 718 Days
• CPC Class Codes: G06F 13/24 using interrupt G06F13/32 takes precedence
• CPC Class Codes: H04L 12/4633 Interconnection of networks using encapsulation techniques, e.g. tunneling
• CPC Class Codes: H04L 45/7453 using hashing
• CPC Class Codes: H04L 47/6295 using multiple queues, one for each individual QoS, connection, flow or priority
• CPC Class Codes: H04L 49/9068 in the network interface card
• CPC Class Codes: H04L 67/02 based on web technology, e.g. hypertext transfer protocol [HTTP]
• CPC Class Codes: H04L 67/60 Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources admission control or resource allocation H04L47/70
• 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 69/163 In-band adaptation of TCP data exchange; In-band control procedures
• CPC Class Codes: H04L 69/22 Parsing or analysis of headers

Abstract

Systems, methods, and other embodiments that interrupt multiple processors to process packets of a single flow are described. In one embodiment, a method includes maintaining a plurality of queues to hold interrupt requests created for packets received by a network device. Each queue is associated with a processor that will be interrupted by interrupt requests to process packets. The network device is configured to control how to queue interrupt requests created for a flow of packets. The network device is configured such that when a threshold number of interrupt requests created for the flow of packets have been inserted within a queue, subsequent interrupt requests created for subsequently received packets of the flow are inserted in a different queue. In this way, the network device is controlled to interrupt multiple processors for the flow of packets by queuing interrupt requests for the flow into different queues.

Claims

1. A non-transitory computer-readable medium storing computer-executable instructions that when executed by a processor of a computing device causes the processor to:

maintain a plurality of queues configured to hold interrupt requests created in response to packets received by a network device, wherein the interrupt requests interrupt a plurality of processors to process the packets, and wherein each queue is associated with a corresponding processor;

configure the network device to queue interrupt requests for a single flow of packets between a source and a destination, wherein the network device is configured to switch from inserting interrupt requests into a first queue to inserting into a second queue upon queuing a threshold number of interrupt requests for the single flow of packets into the first queue;

in response to receiving a packet from a remote computer, evaluate the packet to determine that the packet is part of the single flow between the source as the remote computer and a destination as a service being accessed by the remote computer;

determine a count of interrupt requests inserted into the first queue for the single flow;

responsive to the current count of interrupt requests being less than the threshold number of interrupt requests;

(i) construct an interrupt request for the packet; and

(ii) queue the interrupt request into the first queue; and

control the network device to interrupt multiple processors for the single flow of packets by inserting the interrupt requests for the single flow of packets into different queues of the plurality of queues.

2. The non-transitory computer-readable medium of claim 1, wherein the instructions comprise instructions to:

responsive to queuing the interrupt request into the first queue, increment the count of interrupt requests inserted into the first queue for the single flow.

3. The non-transitory computer-readable medium of claim 1, wherein the instructions comprise instructions to:

determine a count of interrupt requests inserted into the first queue for the single flow;

responsive to the count of interrupt requests triggering the threshold number of interrupt requests,

select a different queue from the plurality of queues for queueing interrupt requests for the single flow;

construct an interrupt request for the packet; and

queue the interrupt request into the different queue from the plurality of queues.

4. The non-transitory computer-readable medium of claim 3, wherein the instructions comprise instructions to:

responsive to queuing the interrupt request into the different queue;

reset the count of interrupt requests to indicate that the interrupt request was inserted into the different queue; and

increment the count of interrupt requests in response to subsequent interrupt requests for the single flow being inserted into the different queue.

5. The non-transitory computer-readable medium of claim 1, wherein the network device is a network card of a computing device, and wherein the instructions comprise instructions to:

control the network card to queue interrupt requests into the queues.

6. The non-transitory computer-readable medium of claim 1, wherein the single flow is a transmission control protocol (TCP) flow of a TCP connection between a remote client device and a service hosted by the computing device, and wherein the instructions comprise instructions to:

construct an interrupt request from a TCP packet of the TCP flow; and

queue the interrupt request.

7. A computing system, comprising:

a processor connected to memory; and

a module stored on a non-transitory computer readable medium and configured with instructions that when executed by the processor cause the processor to;

maintain a plurality of queues configured to hold interrupt requests created in response to packets received by a network device, wherein the interrupt requests interrupt a plurality of processors to process the packets, and wherein each queue is associated with a corresponding processor;

configure the network device to queue interrupt requests for a single flow of packets between a source and a destination, wherein the network device is configured to switch from inserting interrupt requests into a first queue to inserting into a second queue upon queuing a threshold number of interrupt requests for the single flow of packets into the first queue;

in response to receiving a tunnel packet comprising an overlay packet and an underlay packet;

(i) parse the underlay packet to determine that the tunnel packet corresponds to the single flow between the source and the destination; and

(ii) queue the tunnel packet; and

control the network device to interrupt multiple processors for the single flow of packets by inserting interrupt requests for the single flow of packets into different queues of the plurality of queues.

8. The computing system of claim 7, wherein the instructions comprise instructions that cause the processor to:

bypass parsing the overlay packet for determining whether the tunnel packet corresponds to the single flow or a different flow.

9. The computing system of claim 7, wherein the instructions to parse comprise instructions that cause the processor to:

parse a header of the underlay packet to identify a source and destination of the tunnel packet based upon a source internet protocol (IP) address and port and a destination IP and port specified within the header.

10. The computing system of claim 7, wherein interrupt requests for the single flow of packets are queued into multiple queues of the plurality of queues for parallel processing the single flow of packets by multiple processors of the plurality of processors.

11. The computing system of claim 7, wherein the single flow is a user datagram protocol (UDP) flow, and wherein the instructions comprise instructions to:

construct an interrupt request from a UDP packet of the UDP flow; and

queue the interrupt request.

12. A computer-implemented method, the computer-implemented method involving a computing device comprising a processor, and the computer-implemented method comprising:

maintaining, by the processor, a plurality of queues configured to hold interrupt requests created in response to packets received by a network device, wherein the interrupt requests interrupt a plurality of processors to process the packets, and wherein each queue is associated with a corresponding processor;

configuring, by the processor, the network device to queue interrupt requests for a single flow of packets between a source and a destination, wherein the network device is configured to switch from inserting interrupt requests into a first queue to inserting into a second queue upon queuing a threshold number of interrupt requests for the single flow of packets into the first queue;

in response to receiving a packet from a remote computer, evaluating the packet to determine that the packet is part of the single flow of packets between the source as the remote computer and a destination as a service being accessed by the remote computer;

determining a count of interrupt requests inserted into the first queue for the single flow of packets;

responsive to the count of interrupt requests triggering the threshold number of interrupt requests;

(i) selecting a different queue from the plurality of queues for queueing the interrupt requests for the single flow of packets;

(ii) constructing an interrupt request for the packet received; and

(iii) queuing the interrupt request into the different queue of the plurality of queues; and

controlling, by the processor, the network device to interrupt multiple processors for the single flow of packets by inserting the interrupt requests for the single flow of packets into different queues of the plurality of queues.

13. The computer-implemented method of claim 12, further comprising:

configuring a driver of the destination of the single flow to evaluate the plurality of queues for interrupt requests targeting the destination.

14. The computer-implemented method of claim 13, further comprising:

configuring the driver to reorder packets, associated with interrupt requests of the single flow within different queues, for processing by the destination.

15. The computer-implemented method of claim 12, further comprising:

providing each device driver of each processor with access to interrupt requests within corresponding queues associated with each processor.

16. The computer-implemented method of claim 12, further comprising:

receiving a packet from a remote computer;

evaluating the packet to determine that the packet is part of the single flow between the source as the remote computer and a destination as a service being accessed by the remote computer;

determining a count of interrupt requests inserted into the first queue for the single flow;

responsive to the current count of interrupt requests being less than the threshold number of interrupt requests,

constructing an interrupt request for the packet; and

queueing the interrupt request into the first queue.