Reporting elephant flows to a network controller

• US 10,158,538 B2
• Filed: 03/31/2014
• Issued: 12/18/2018
• Est. Priority Date: 12/09/2013
• Status: Active Grant

• Current Assignee: Nicira Incorporated (Broadcom, Inc.)
• Original Assignee: Nicira Incorporated (Broadcom, Inc.)
• Inventors: Pettit, Justin, Casado, Martin, Koponen, Teemu, Davie, Bruce, Lambeth, W. Andrew
• Primary Examiner(s): Jaroenchonwanit, Bunjob
• Application Number:
  US14/231,654
• Publication Number:
  US 20150163145A1
• Time in Patent Office: 1,723 Days
• Field of Search: None
• CPC Class Codes: H04L 43/026 using flow identification
• CPC Class Codes: H04L 43/16 Threshold monitoring
• CPC Class Codes: H04L 47/11 Identifying congestion
• CPC Class Codes: H04L 47/125 by balancing the load, e.g. traffic engineering
• CPC Class Codes: H04L 49/252 Store and forward routing
• CPC Class Codes: H04L 49/3009 Header conversion, routing tables or routing tags
• CPC Class Codes: H04L 49/3018 Input queuing
• CPC Class Codes: H04L 67/10 in which an application is distributed across nodes in the network software deployment G06F8/60; multiprogramming arrangements G06F9/46

Abstract

Some embodiments provide a system that detects whether a flow is an elephant flow; and if so, the system treats it differently than a mouse flow. The system of some embodiment detect elephants based on one or more of the following: statistics associated with a flow, packet segment size, and invoked system calls. Also, some embodiments use one or more various methods to handle elephant flows. Examples of such methods include marking each packet belonging to an elephant with a particular marking, breaking the elephants into mice, reporting the elephant to a network controller, and selectively choosing a route for each packet belonging to the elephant.

Claims

1. A non-transitory machine readable medium storing a program, that when executed by at least one processing unit of a host computer implements an edge forwarding element on the host computer, the program comprising sets of instructions for:

monitoring data flows sent from a plurality of machines that execute on the host computer along with the edge forwarding element;

generating a set of statistics related to each monitored data flow;

determining that one of the data flows is an elephant flow based on the generated statistics;

reporting the elephant flow to a set of network controllers;

receiving, from the set of network controllers, configuration data that configures the edge forwarding element to break the elephant flow into a plurality of mice flows that are sent along different paths to a destination of the elephant flow; and

using equal cost multipath (ECMP) routing to break the elephant flow into the plurality of mice flows and sending each mice flow along one of the plurality of paths to the elephant flow'"'"'s destination.

2. The non-transitory machine readable medium of claim 1, wherein the edge forwarding element is part of a network that comprises edge and non-edge forwarding elements, and the non-edge forwarding elements comprising hardware top of the rack switches.

3. The non-transitory machine readable medium of claim 2, wherein the program further comprises a set of instructions for establishing, based on a tunnel configuration received from the set of network controllers, a tunnel between the edge forwarding element of the host computer and another edge forwarding element of another host computer.

4. The non-transitory machine readable medium of claim 1, wherein the configuration data from the set of network controllers configures the edge forwarding element to use ECMP routing to break the elephant flow into the plurality of mice flows and to send each mice flow along one of the plurality of paths to the elephant flow'"'"'s destination.

5. The non-transitory machine readable medium of claim 4, wherein the set of instructions for sending each mice flow along one of the plurality of paths comprises sending a first mice flow through a first egress port and sending a second mice flow through a second egress port.

6. A method for an edge forwarding element that operates on a host computer, the edge forwarding element part of a network that comprises edge and non-edge forwarding elements, the method comprising:

at the edge forwarding element, monitoring data flows sent from a plurality of machines that execute on the host computer along with the edge forwarding element;

generating a set of statistics related to each monitored data flow;

determining, at the edge forwarding element, that one of the data flows is an elephant flow based on the generated statistics;

sending a report identifying the elephant flow to a network controller that, based on the report, configures one particular non-edge forwarding element to break the elephant flow into a plurality of mice flows; and

forwarding the packets of the elephant flow through the network, wherein when the particular non-edge forwarding element receives the elephant flow packets, the particular non-edge forwarding element breaks the elephant flow into the plurality of mice flows and sends each mice flow along a different one of a plurality of paths to a destination of the elephant flow.

7. The method of claim 6, wherein, based on the report identifying the elephant flow, the network controller further configures a set of edge forwarding elements.

8. The method of claim 7, wherein the non-edge forwarding elements operate as software forwarding elements on a computing device.

9. The method of claim 6, wherein the non-edge forwarding elements of the network comprise hardware top of the rack switches.

10. The method of claim 6, wherein the network controller configures the particular non-edge forwarding element to use equal cost multipath (ECMP) routing to break the elephant flow into a plurality of mice flows.

11. The method of claim 10, wherein the particular non-edge forwarding element sends a first mice flow through a first egress port and sending a second mice flow through a second egress port.

12. A non-transitory machine readable medium storing a program, that when executed by at least one processing unit of a host computer implements an edge forwarding element on the host computer, the edge forwarding element part of a network that comprises edge and non-edge forwarding elements, the program comprising sets of instructions for:

monitoring data flows sent from a plurality of machines that execute on the host computer along with the edge forwarding element;

generating a set of statistics related to each monitored data flow;

determining that one of the data flows is an elephant flow based on the generated statistics;

sending a report identifying the elephant flow to a network controller that, based on the report, configures one particular non-edge forwarding element to break the elephant flow into a plurality of mice flows; and

forwarding the packets of the elephant flow through the network, wherein when the particular non-edge forwarding element receives the elephant flow packets, the particular non-edge forwarding element breaks the elephant flow into the plurality of mice flows and sends each mice flow along a different one of a plurality of paths to a destination of the elephant flow.

13. The non-transitory machine readable medium of claim 12, wherein the network controller configures the particular non-edge forwarding element to perform different quality of service (QoS) operations for the elephant flow than other data flows.

14. The non-transitory machine readable medium of claim 12, wherein the network controller provides configuration data to the particular non-edge forwarding element that directs the particular non-edge forwarding element to perform equal cost multipath (ECMP) routing to break the elephant flow into the plurality of mice flows sent along one of the plurality of paths to the elephant flow'"'"'s destination.

15. A non-transitory machine readable medium storing a program, that when executed by at least one processing unit of a host computer implements an edge forwarding element on the host computer, the edge forwarding element part of a network that comprises edge and non-edge forwarding elements, the program comprising sets of instructions for:

monitoring data flows sent from a plurality of machines that execute on the host computer along with the edge forwarding element;

generating a set of statistics related to each monitored data flow;

determining that one of the data flows is an elephant flow based on the generated statistics;

sending a report comprising encapsulation header information to the network controller identifying the elephant flow, wherein the controller, based on the report, configures one particular non-edge forwarding element to break the elephant flow into the plurality of mice flows based on the encapsulation header information; and

encapsulating the monitored data flows and forwarding the packets of the elephant flow through the network, wherein when the particular non-edge forwarding element receives the elephant flow packets, the particular non-edge forwarding element breaks the elephant flow into the plurality of mice flows and sends each mice flow along a different one of a plurality of paths to a destination of the elephant flow.

16. The non-transitory machine readable medium of claim 15, wherein the network controller provides configuration data to the particular non-edge forwarding element that directs the particular non-edge forwarding element to perform equal cost multipath (ECMP ) routing to break the elephant flow into the plurality of mice flows sent along one of the plurality of paths to the elephant flow'"'"'s destination.