Congestion notification in leaf and spine networks

Patent metadata

Field	Value
Publication number	US-9246818-B2
Application number	US-201313924950-A
Country	US
Kind code	B2
Filing date	Jun 24, 2013
Priority date	Jun 24, 2013
Publication date	Jan 26, 2016
Grant date	Jan 26, 2016

Abstract

Methods and systems to alleviate congestion in leaf and spine network topologies. Control information may be advertised by spines regarding the leaves to which it is connected. The control information is advertised to other leaf switches connected to the advertising spine. The control information may include a destination leaf ID, an enroute spine ID, an FL flag and an FF flag. The FL flag is a “Flow Learn Flag,” that is set by the advertising spine that a leaf switch toward which the spine may be facing congestion above at a first predetermined threshold level. The FF flag is a “Flow Freeze Flag,” that is set when all links to the corresponding destination leaf identified are congested, and the congestion level has crossed a second predetermined threshold level.

First claim

What is claimed is:

1. A method comprising:
receiving control information generated by a first network device at a second network device, wherein a control packet comprises a first flag set in accordance with a total of all path capacity or link congestion being greater than a first predetermined threshold level and second flag set in accordance with the total of all path capacity or link congestion being greater than a second predetermined level;
determining a destination location specified in control information in accordance with a MAC address table, Routing information Table or Switching Information Table;
determining if the first flag is set, if the first flag is set and the second flag not set, and if so:
creating an entry in a flow table; and
forwarding to the destination location using a spine node identified in the control information;

determining if the second flag is set, and if so:
forwarding to the destination using the flow entry if available for that traffic, otherwise forwarding by avoiding a spine node identified in the control information; and

determining if none of the first flag and the second flag is set, and if so forwarding normally.

2. The method of claim 1 , if the second flag is set, the method further comprising checking a flow table for a matching entry to the destination entry, and if there is no match, performing load balancing to exclude the spine node identified in the control information with the second flag set.

3. The method of claim 1 , further comprising maintaining the entry in the flow table until the link congestion falls below a predetermined reset threshold level.

4. The method of claim 1 , wherein the method is implemented in a leaf and spine network, the method further comprising:
receiving the control information at each leaf in the network; and
learning flows from a first server to a second server in accordance with the first flag.

5. The method of claim 1 , further comprising setting the first predetermined threshold level to 70% of a total of all path capacity.

6. The method of claim 1 , further comprising setting the second predetermined threshold level to 90% of a total of all path capacity.

7. The method of claim 1 , wherein a spine advertises the control information regarding the leaves to which it is connected to other leaf switches connected to the spine.

8. The method of claim 1 , further comprising embedding the control information in a protocol used to exchange topological information in a network.

9. A non-transitory computer readable storage media encoded with software comprising computer executable instructions and when the software is executed operable to:
receive control information generated by a first network device at a second network device, wherein a control packet comprises a first flag set in accordance with a total of all path capacity or link congestion being greater than a first predetermined threshold level and second flag set in accordance with the total of all path capacity or link congestion being greater than a second predetermined level;
determine a destination location specified in control information in accordance with a MAC address table, Routing information Table or Switching Information Table;
determine if the first flag is set, if the first flag is set and the second flag not set, and if so:
create an entry in a flow table; and
forward to the destination location using a spine node identified in the control information;

determine if the second flag is set, and if so:
forward to the destination using the flow entry if available for that traffic, otherwise forward by avoiding a spine node identified in the control information; and

determine if none of the first flag and the second flag is set, and if so forward normally.

10. The non-transitory computer readable storage media of claim 9 , if the second flag is set, further comprising instructions to check a flow table for a matching entry to the destination entry, and if there is no match, performing load balancing to exclude the spine node identified in the control information with the second flag set.

11. The non-transitory computer readable storage media of claim 9 , further comprising instructions to maintain the entry in the flow table until the link congestion falls below a predetermined reset threshold level.

12. The non-transitory computer readable storage media of claim 9 , wherein in a leaf and spine network, further comprising instructions to:
receive the control information at each leaf in the network; and
learn flows from a first server to a second server in accordance with the first flag.

13. The non-transitory computer readable storage media of claim 9 , wherein the first predetermined threshold level is set to 70% of a total of all path capacity.

14. The non-transitory computer readable storage media of claim 9 , wherein the second predetermined threshold level is set to 90% of a total of all path capacity.

15. The non-transitory computer readable storage media of claim 9 , wherein a spine advertises the control information regarding the leaves to which it is connected to other leaf switches connected to the spine.

16. The non-transitory computer readable storage media of claim 9 , further comprising instructions to embed the control information in a protocol used to exchange topological information in a network.

17. An apparatus comprising:
a plurality of network interfaces;
a memory; and
a processor coupled to the network interfaces and memory, and configured to:
receive control information generated by a first network device at a second network device, wherein a control packet comprises a first flag set in accordance with a total of all path capacity or link congestion being greater than a first predetermined threshold level and second flag set in accordance with the total of all path capacity or link congestion being greater than a second predetermined level;
determine a destination location specified in control information in accordance with a MAC address table, Routing information Table or Switching Information Table;
determine if the first flag is set, if the first flag is set and the second flag not set, and if so:
create an entry in a flow table; and
forward to the destination location using a spine node identified in the control information;

determine if the second flag is set, and if so:
forward to the destination using the flow entry if available for that traffic, otherwise forward by avoiding a spine node identified in the control information; and

determine if none of the first flag and the second flag is set, and if so forward normally.

18. The apparatus of claim 17 , the processor further configured to maintain the entry in the flow table until the link congestion falls below a predetermined reset threshold level.

19. The apparatus of claim 17 , if the second flag is set, the processor is further configured to check a flow table for a matching entry to the destination entry, and if there is no match, performing load balancing to exclude the spine node identified in the control information with the second flag set.

20. The apparatus of claim 17 , wherein in a leaf and spine network, the processor is further configured to:
receive the control information at each leaf in the network; and
learn flows from a first server to a second server in accordance with the first flag.

Assignees

• Cisco Tech Inc

Inventors

• Deshpande Sushrut Sudhakar
• Lele Aniket

Classifications

• H04L47/12 Primary

  Avoiding congestion; Recovering from congestion · CPC title