System and method for detection of elephant flows

1. A switching device comprising:

one or more ports;

a memory; and

a control unit coupled to the ports and the memory, the control unit being configured to:

detect an elephant flow between a computing device and a storage node;

receive one or more first network packets at the switching device;

determine whether the first network packets are associated with the elephant flow; and

forward the first network packets as elephant flow packets when the first network packets are associated with the elephant flow;

wherein to detect the elephant flow, the control unit is configured to:

detect storage locations information included in one or more first messages received for forwarding to the computing device on one or more of the ports, the storage locations information including an address of the storage node and one or more storage locations within the storage node;

detect subsequent opening of a connection between the computing device and the storage node based on one or more second messages received for forwarding on one or more of the ports; and

determine as identifying characteristics of the elephant flow, one or more properties that identify the connection;

wherein to determine whether the first network packets are associated with the elephant flow, the control unit is further configured to determine whether characteristics of the first network packets match the identifying characteristics of the elephant flow.

2. The switching device of claim 1 , wherein the control unit is further configured to store the identifying characteristics in the memory.

3. The switching device of claim 1 , wherein the control unit is further configured to alter a forwarding strategy used by the switching device to forward the first network packets associated with the elephant flow.

4. The switching device of claim 3 , wherein the control unit is further configured to:

receive the first network packets on one or more of the ports; and

forward the first network packets on one or more of the ports using the altered forwarding strategy.

5. The switching device of claim 3 , wherein the control unit is further configured to alter the forwarding strategy by performing one or more tasks selected from a group consisting of increasing a bandwidth allocated to the first network packets, finding one or more first alternate paths for the first network packets, finding one or more second alternate paths for second network packets not associated with the elephant flow, implementing one or more quality of service (QoS) mechanisms for the first network packets, and activating one or more QoS mechanisms for the first network packets.

6. The switching device of claim 3 , wherein the control unit is further configured to detect a closing of the connection.

7. The switching device of claim 6 , wherein the control unit is further configured to undo the altering of the forwarding strategy of the switching device based on the closing of the connection.

8. The switching device of claim 1 , wherein the control unit is further configured to notify one or more other switching devices in a network of the determined identifying characteristics of the elephant flow.

9. The switching device of claim 8 , wherein the control unit is further configured to notify a network controller of the determined identifying characteristics of the elephant flow.

10. The switching device of claim 1 , wherein:

the connection is a transport control protocol (TCP) connection; and

the one or more properties that identify the connection include an internet protocol (IP) address of the computing device, an IP address of the storage node, a TCP port associated with the computing device, and a TCP port associated with the storage node.

11. The switching device of claim 1 , wherein the computing device is a client or a first data node and the storage node is a second data node.

12. The switching device of claim 1 , wherein one or more of the first messages are forwarded from a name node.

13. The switching device of claim 1 , wherein, the switching device is configured to couple the computing device to a network.

14. A method of managing a switching device, the method comprising:

detecting an elephant flow by detecting storage locations information included in one or more first messages received for forwarding by the switching device to a computing device, the storage locations information including an address of a storage node and one or more storage locations within the storage node, detecting subsequent opening of a connection between the computing device and the storage node based on one or more second messages received for forwarding at the switching device, and determining as identifying characteristics of the elephant flow, one or more properties that identify the connection;

receiving one or more network packets at the switching device;

determining whether the network packets are associated with the elephant flow based on the identifying characteristics; and

forwarding the network packets using an altered forwarding strategy when the network packets are associated with the elephant flow;

wherein the first and second messages and the network packets are received on ports of the switching device.

15. The method of claim 14 , further comprising sending the identifying characteristics to one or more other switching devices in a network.

16. The method of claim 14 , further comprising sending the identifying characteristics to a network controller.

17. The method of claim 14 , further comprising:

detecting closing of the connection; and

forwarding the network packets without using the altered forwarding strategy after the connection is closed.

18. An information handling system comprising:

a switching device comprising:

one or more ports;

a memory; and

a control unit coupled to the ports and the memory;

wherein the control unit is configured to:

detect an elephant flow by detecting storage locations information included in one or more first messages received, on one or more of the ports, for forwarding, the first messages being exchanged between a Hadoop client on a computing device and a name node of a Hadoop distributed file system (HDFS) or the Hadoop client and a first data node of the HDFS, and the storage locations information including an address of a second data node of the HDFS and one or more storage locations within the second data node, detecting subsequent opening of a connection between the Hadoop client or a storage module of the first data node and a storage module of the second data node based on one or more second messages received for forwarding on one or more of the ports, and determining as identifying characteristics of the elephant flow, one or more properties that identify the connection;

determine whether one or more network packets received for forwarding on one or more of the ports are associated with the elephant flow based on the identifying characteristics; and

forward the network packets using an altered forwarding strategy when the network packets are associated with the elephant flow.

19. The information handling system of claim 18 , wherein:

the connection is a transport control protocol (TCP) connection; and

the one or more properties that identify the connection include an internet protocol (IP) address of the computing device or the first data node, an IP address of the second data node, a TCP port associated with the Hadoop client or the storage module of the first data node, and a TCP port associated with the storage module of the second data node.