Decoupled and concurrent packet processing and packet buffering for load-balancing router architecture

• US 9,014,003 B2
• Filed: 11/08/2011
• Issued: 04/21/2015
• Est. Priority Date: 11/08/2011
• Status: Active Grant

• Current Assignee: Futurewei Technologies Incorporated (Huawei Investment & Holding Co., Ltd.)
• Original Assignee: Futurewei Technologies Incorporated (Huawei Investment & Holding Co., Ltd.)
• Inventors: Song, Haoyu, Boures, Thomas J.
• Primary Examiner(s): Hsu, Alpus H
• Assistant Examiner(s): PATEL, DHARMESH J
• Application Number:
  US13/291,195
• Publication Number:
  US 20130114413A1
• Time in Patent Office: 1,260 Days
• Field of Search: 370/235, 370/237, 370/392, 370/395.32, 370/409, 709/223, 709/226, 709/242, 709/245, 709/249
• US Class Current: 370/237
• CPC Class Codes: H04L 41/0896 Bandwidth or capacity management, i.e. automatically increasing or decreasing capacities flow or congestion control using dynamic resource allocation, e.g. in-call renegotiation, H04L47/76
• CPC Class Codes: H04L 47/10 Flow control; Congestion control
• CPC Class Codes: H04L 47/43 Assembling or disassembling of packets, e.g. segmentation and reassembly [SAR]
• CPC Class Codes: H04L 49/9042 Separate storage for different parts of the packet, e.g. header and payload

Abstract

A network component comprising a traffic manager (TM) configured to assign a unique identifier (ID) to a received packet, a splitter module configured to separate a header and a payload of the received packet, the splitter module further configured to attach the unique ID to each of the header and the payload, and a merger module configured to assemble an outgoing packet using the header and the payload.

Claims

1. A network component comprising:

a traffic manager (TM) configured to assign a unique identifier (ID) to a received packet;

a splitter module configured to separate a header and a payload of the received packet, attach the unique ID to the header, and attach the unique ID to the payload;

a merger module configured to receive the header from a first network processing unit (NPU) of a plurality of NPUs, obtain the unique ID attached to the header, obtain the payload from a first buffer that holds the payload, wherein the unique ID attached to the header is used to determine that the first buffer holds the payload, and merge the header and the payload according to the unique ID to form an outgoing packet;

a packet buffer load balancer configured to receive the payload from the splitter module after attaching the unique ID to the payload, determine the first buffer amongst a plurality of packet buffers to hold the payload, and determine that the first buffer holds the payload based upon a first predefined policy for maximizing memory bandwidth and capacity utilization while minimizing memory access latency; and

a packet processing load balancer configured to receive the header from the splitter module after attaching the unique ID to the header, determine the first NPU amongst the plurality of NPUs to process the header, and determine that the header is processed by the first NPU based upon a second predefined policy for maximizing throughput while minimizing a processing latency,

wherein the header and the payload of the received packet are processed in parallel with each other prior to merging the header and the payload to form the outgoing packet.

2. The network component of claim 1, wherein the packet buffer load balancer is further configured to distribute the payload to the first buffer amongst the plurality of buffers, wherein the packet processing load balancer is configured to distribute the header to the first NPU amongst the plurality of NPUs.

3. The network component of claim 2 further comprising a lookup table configured to store a location of the payload following attaching the unique ID to the payload, wherein the merger module is further configured to search the lookup table using the unique ID attached to the header as a search key for the lookup table in order to retrieve the location of the payload.

4. The network component of claim 3, wherein the location of the payload is an address of the first buffer, and wherein the merger module is further configured to receive the payload from the first buffer after obtaining the address of the first buffer from the lookup table.

5. The network component of claim 4, wherein the lookup table does not release an entry after receiving the payload, and wherein the merger module is further configured to subsequently re-read the payload from the first buffer.

6. The network component of claim 2, wherein a first quantity of the plurality of NPUs is different from a second quantity of the plurality of packet buffers.

7. In a traffic manager (TM), a method for load-balancing, the method comprising:

receiving a packet;

assigning a unique identifier (ID) to the packet;

splitting the packet into a header and a payload for separate processing of the header and the payload;

attaching the unique ID to the header;

attaching the unique ID to the payload;

assigning the header to one of a plurality of network processing units (NPUs) for processing the header after splitting the packet, wherein the assigning the header to the one of the NPUs is determined by a first predefined policy for maximizing throughput while minimizing latency;

assigning the payload to one of a plurality of packet buffers, wherein assigning the payload to the one of the buffers is determined by a second predefined policy for maximizing memory bandwidth and capacity utilization while minimizing memory access latency;

receiving the header from the one of a plurality of NPUs;

determining the unique ID attached to the header;

retrieving the payload located within the one of a plurality of buffers, wherein the one of the buffers is identified using the unique ID attached to the header; and

subsequently reassembling the payload and the header to form an outgoing packet using the unique ID,

wherein the header and the payload are processed in parallel with each other prior to reassembling the header and the payload to form the outgoing packet.

8. The method of claim 7 further comprising:

attaching the unique ID to the header prior to processing the header at the one of the NPUs; and

attaching the unique ID to the payload prior to buffering the payload at the one of the buffers.

9. The method of claim 8 further comprising storing a location of the payload in a lookup table that references the one of the buffers following attaching the unique ID to the payload.

10. The method of claim 9 further comprising:

processing the header by the one of the NPUs;

retrieving the unique ID attached to the header;

performing a lookup using the unique ID as a search key to determine the location of the payload; and

reading the payload from the one of the buffers in response to performing the lookup.

11. The method of claim 10, wherein the location is a memory address, and wherein the method further comprises directing that the memory address not be released after reading the payload and subsequently re-reading the payload from the one of the buffers.

12. The method of claim 8, wherein a quantity of the plurality of NPUs is different than a quantity of the plurality of packet buffers, and wherein assigning the header and assigning the payload occurs in parallel.

13. The method of claim 7, further comprising selectively powering down one or more of a plurality of NPUs or selectively powering up the one or more of the plurality of NPUs based upon actual or expected network traffic complexity.

14. The method of claim 7, further comprising selectively powering down one or more of a plurality of packet buffers or selectively powering up the one or more of the plurality of packet buffers based upon actual or expected network traffic volume.

15. A computer program product for load-balancing data packets, the computer program product comprising computer executable instructions stored in a non-transitory computer readable medium such that when executed by a processor causes a network component to:

assign a unique identifier (ID) to a received data packet;

separate a header and a payload of the received data packet;

attach the unique ID to both the header and the payload;

distribute the header with the attached unique ID to one of a plurality of network processing units (NPUs) according to a predefined load balancing policy;

distribute the payload with the attached unique ID to one of a plurality of buffers according to a second predefined load balancing policy;

receiving the header from the one of a plurality of NPUs;

determining the unique ID attached to the header;

obtain the payload from the one of a plurality of buffers referenced by the unique ID attached to the header; and

merge the header and the payload after obtaining the payload from the one of the buffers,

wherein the header and the payload of the received data packet are processed in parallel with each other prior to assembling the header and the payload to form an outgoing packet.

16. The computer program product of claim 15, wherein the instructions, when executed by a processor, further causes a network component to:

attach the unique ID to both the header and the payload prior to buffering the payload in the one of the buffers;

store an address of the one of the buffers used to buffer the payload in a lookup table following attaching the unique ID to the payload;

map the address of one of the buffers used to buffer in the lookup table to the unique ID; and

retrieve the address of the one of the buffers used to buffer the payload from the lookup table using the unique ID of the header as a search key for the lookup table.

17. The computer program product of claim 15,

wherein the predefined load balancing policy selects the one of the NPUs from a packet rate, and

wherein the second predefined load balancing policy selects the one of the buffers from a bit rate.