TCP offload device that load balances and fails-over between aggregated ports having different MAC addresses

• US 20030140124A1
• Filed: 08/27/2002
• Published: 07/24/2003
• Est. Priority Date: 03/07/2001
• Status: Active Grant

• Current Assignee: Alacritech, Inc.
• Original Assignee: Alacritech, Inc.
• Inventors: Burns, David A.
• Granted Patent: US 6,938,092 B2
• Time in Patent Office: Days
• US Class Current: 709/220
• CPC Class Codes: H04L 1/16 in which the return channel carries supervisory signals, e.g. repetition request signals
• CPC Class Codes: H04L 45/00 Routing or path finding of packets in data switching networks routing or path finding in wireless networks H04W40/00
• CPC Class Codes: H04L 45/245 Link aggregation, e.g. trunking
• CPC Class Codes: H04L 49/90 Buffering arrangements
• CPC Class Codes: H04L 49/9063 Intermediate storage in different physical parts of a node or terminal
• 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 69/165 Combined use of TCP and UDP protocols; selection criteria therefor
• CPC Class Codes: H04L 69/168 specially adapted for link layer protocols, e.g. asynchronous transfer mode [ATM], synchronous optical network [SONET] or point-to-point protocol [PPP]
• CPC Class Codes: Y02D 30/50 in wire-line communication networks, e.g. low power modes or reduced link rate

Abstract

TCP/IP traffic passing over a TCP connection is initially load balanced by a TCP offload network interface device (NID) in fast-path between multiple ports of an aggregation team. A failure then occurs on one of the ports. Processing of the TCP connection then switches to slow-path. The function of the failed port is then assumed by another functioning port of the team such that fast-path processing of the TCP connection by the TCP offload NID automatically resumes. The two ports of the TCP offload NID (the failed port and the port that assumes its functionality) use different MAC addresses. Fast-path load balancing and fail-over are therefore possible without the use of a Fast-Etherchannel switch that would have required the two ports to use the same MAC address.

Claims

1. A method, comprising:

performing TCP protocol processing on a TCP offload device such that a host coupled to the TCP offload device is offloaded of substantially all TCP protocol processing tasks associated with communicating network information across a TCP connection between the TCP offload device and a second device, a first port of the TCP offload device being identified by a first MAC address, a second port of the TCP offload device being identified by a second MAC address, the first port of the TCP offload device being used to communicate said network information between the TCP offload device and the second device; and

after a communication failure associated with the first port, switching from using the first port to communicate said network information between the TCP offload device and the second device to using the second port to communicate said network information between the TCP offload device and the second device, wherein after said switching the second port of the TCP offload device is identified by the first MAC address, wherein after said switching the TCP offload device performs substantially all TCP protocol processing associated with communicating said network information across the TCP connection via the second port.

2. The method of claim 1, wherein the TCP offload device is operable in a fast-path mode and in a slow-path mode, the host executing a protocol stack involving a plurality of protocol processing layers,

wherein in the slow-path mode the stack of the host performs substantial TCP protocol processing on network information communicated across the TCP connection, and

wherein in the fast-path mode the TCP offload device performs substantially all TCP protocol processing associated with communicating across the TCP connection such that the stack of the host performs substantially no TCP protocol processing for the TCP connection when in the fast-path mode.

3. The method of claim 2, wherein after said communication failure the TCP offload device switches from fast-path processing of the TCP connection to slow-path processing of the TCP connection, and wherein the TCP offload device then switches back to fast-path processing of the TCP connection such that the second port communicates said network information when the TCP offload device is in the fast-path mode.

4. The method of claim 1, wherein a block of information is maintained for the TCP connection, the block storing a state of the TCP connection, the block including MAC source address information for the TCP connection, the method further comprising:

replacing the MAC source address information of the block with the first MAC address, the replacing occuring after the communication failure and before the switching from the first port.

5. The method of claim 4, wherein the block of information is a transmit control block (TCB), wherein before said communication failure the second port was used to transmit network information for the TCP connection from the TCP offload device, and wherein after said communication failure the first port was used to transmit network information for the TCP connection from the TCP offload device.

6. The method of claim 1, wherein the second device is taken from the group consisting of:

a switch and a hub.

7. The method of claim 1, wherein a first portion of said network information is communicated across the TCP connection between the TCP offload device and the second device before the communication failure, and wherein a second portion of said network information is communicated across the TCP connection between the TCP offload device and the second device after the communication failure.

8. The method of claim 1, wherein the network information is a TCP message.

9. The method of claim 1, the method further comprising:

transmitting an ARP request from the TCP offload device, the ARP request indicating that the first MAC address is associated with the second port, the ARP request being transmitted from the TCP offload device after the communication failure and before the switching.

10. The method of claim 1, wherein the network information communicated between the TCP offload device and the second device includes a first packet and a plurality of subsequent packets, and wherein the host performs TCP protocol processing on the first packet, and wherein the host performs no TCP protocol processing on the subsequent packets.

11. The method of claim 1, wherein before the communication failure a first portion of network information for the TCP connection is communicated via the first port of the TCP offload device and wherein a second portion of network information for the TCP connection is communicated via the second port of the TCP offload device, and wherein before the communication failure the first port is identified by the first MAC address and the second port is identified by the second MAC address.

12. A method, comprising:

load balancing network traffic between a first port of a TCP offload device and a second port of the TCP offload device, the network traffic being information communicated across a TCP connection between the TCP offload device and a second device, the first port being identified by a first MAC address, the second port being identified by a second MAC address; and

failing over from the first port to the second port such that the network traffic communicated across the TCP connection after the fail over is communicated via the second port and not via the first port, and wherein after the fail over the second port is identified by the first MAC address.

13. The method of claim 12, wherein the TCP offload device is coupled to a host, the host executing a stack of protocol processing layers, wherein during said load balancing said TCP offload device performs substantially all TCP protocol processing of said network traffic such that said stack performs substantially no TCP protocol processing of said network information, and wherein after the fail over said TCP offload device performs substantially all TCP protocol processing of said network traffic such that said stack performs substantially no TCP protocol processing of said network information.

14. The method of claim 13, wherein after the fail over the second port is identified by the first MAC address.

15. A set of instructions executable on a host, the host being coupled to a TCP offload network interface device, the set of instructions being for performing steps comprising:

(a) controlling the TCP offload network interface device such that the TCP offload network interface device fast-path processes network communications communicated between the TCP offload network interface device and a second device, the network communications being communicated across a TCP connection, the network communications being communicated via a first port of the TCP offload network interface device;

(b) maintaining a list of teams, one of the teams including a first team member and a second team member, the first team member identifying the first port of the TCP offload device and having a MAC address field associated with the first port, a first MAC address being in the MAC address field of the first port, the second team member identifying a second port of the TCP offload device and having a MAC address field associated with the second port, a second MAC address being in the MAC address field of the second port;

(c) in response to a communication failure associated with the first port, writing the first MAC address into the MAC address field of the second port; and

(d) controlling the TCP offload network interface device such that the TCP offload network interface device communicates the network communications across the TCP connection via the second port and not via the first port, wherein the TCP offload network interface device fast-path processes the network communications after the failure, and wherein after the failure the second port of the TCP offload network interface device is associated with the first MAC address.

16. The set of instructions of claim 15, wherein the set of instructions controls the TCP offload network interface device in (a) at least in part by passing control of the TCP connection to the TCP offload network interface device, the set of instructions passing control of the TCP connection to the TCP offload network interface device by passing a Transmit Control Block (TCB) to the TCP offload network interface device.

17. The set of instructions of claim 15, wherein the set of instructions controls the TCP offload network interface device in (d) at least in part by passing control of the TCP connection to the TCP offload network interface device, the set of instructions passing control of the TCP connection to the TCP offload network interface device by passing a Transmit Control Block (TCB) to the TCP offload network interface device, the TCB including a template header, the template header having a MAC source address field, the MAC source address field containing the first MAC address.

18. The set of instructions of claim 15, wherein the TCP offload network interface device initially fast-path processes the network communications for the TCP connection in (a), then switches to slow-path processing of the network communications for the TCP connection, and then switches back to fast-path processing the network communications for the TCP connection in (d).

19. The set of instructions of Claim 15, wherein the set of instructions includes a stack of protocol processing layers and a port aggregation driver, the port aggregation driver maintaining the list of teams.

20. A system comprising a network interface device operatively coupled to a host, the system comprising:

a first physical port;

a second physical port; and

means for (a) load balancing first network traffic between the first physical port and the second physical port, the first network traffic being communicated across a single first TCP connection, wherein a first packet of the first TCP connection is communicated through the first physical port, wherein a second packet of the first TCP connection is communicated through the second physical port, and wherein during said load balancing the first physical port is identified by a first MAC address and the second physical port is identified by a second MAC address, and wherein the means is also for (b) failing over second network traffic communicated across a single second TCP connection, the second network traffic being failed over from one of the first and second physical ports to the other, wherein both before and after said fail over the network interface device fast-path processes the second network traffic of the second TCP connection such that the network interface device performs substantially all TCP protocol processing on the second network traffic and such that the host performs substantially no TCP protocol processing on the second network traffic.

21. The system of claim 20, wherein the means comprises the network interface device and a port aggregation driver on the host.

22. The system of claim 20, wherein a communication failure occurs, wherein in response to the failure the network interface device switches from fast-path processing the second network traffic to slow-path processing the second network traffic and passes control of the second TCP connection to the host, wherein the host then performs incidental TCP processing on the second network traffic, wherein the host then passes control of the TCP connection back to the network interface device, and wherein the network interface device then resumes fast-path processing on the second network traffic.