Load balancing in a data delivery system

• US 20030061356A1
• Filed: 09/24/2001
• Published: 03/27/2003
• Est. Priority Date: 09/24/2001
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Jason, James L. Jr.
• Granted Patent: US 7,039,916 B2
• Time in Patent Office: Days
• US Class Current: 709/227
• CPC Class Codes: H04L 43/091 Measuring contribution of individual network components to actual service level
• CPC Class Codes: H04L 67/1001 for accessing one among a plurality of replicated servers
• CPC Class Codes: H04L 67/1008 based on parameters of servers, e.g. available memory or workload monitoring of computer activity G06F11/30
• CPC Class Codes: H04L 67/101 based on network conditions
• CPC Class Codes: H04L 67/1034 Reaction to server failures by a load balancer

Abstract

The time taken for connection establishment is monitored to aid in selecting load distribution among nodes in a data delivery system, such as a server cluster. The failure of a node to respond to a connection request may be used to identify a crashed node. The number of connections being maintained and the amount of bandwidth being consumed may also be monitored for each node, and this information may be used to determine when a node should be removed from contention for new connection requests and when a node should be reinstated to receive new connection requests.

Claims

1. A system comprising:

a communications interface configurable to monitor messages on a network;

a processor coupled with the communications interface; and

a machine-readable medium coupled with the processor, the machine-readable medium being configured to instruct the processor to access machine-readable instructions to cause the processor to perform operations including,

identifying a message requesting a connection with a data delivery system having a plurality of nodes,

initiating a tracking of a duration after the message identification, and

identifying a node of the data delivery system as being overloaded based upon the tracked duration.

2. The system of claim 1, wherein the operations further include balancing load among the plurality of nodes in the data delivery system.

3. The system of claim 2, wherein balancing load among the plurality of nodes comprises removing the overloaded node from contention for new connections.

4. The system of claim 3, wherein removing the overloaded node comprises transmitting a message to initiate a removal.

5. The system of claim 2, wherein the tracked duration comprises an amount of time between the message identification and either an identification of an acknowledging message from the node or a termination of a time-out period during which no acknowledging message is returned by the node.

6. The system of claim 5, wherein the node identifying operation comprises comparing the tracked duration to an historical average for the node.

7. The system of claim 5, wherein the communications interface comprises an Ethernet card.

8. The system of claim 7, wherein the network comprises an Internet Protocol network.

9. The system of claim 8, wherein the data delivery system comprises a web server cluster.

10. A method of adjusting assignment of connection requests to nodes in a data delivery system, the method comprising:

monitoring messages on a network;

identifying a message requesting a connection with a data delivery system having multiple nodes;

initiating a tracking of a duration after the message identification; and

adjusting assignment of connection requests based upon the tracked duration.

11. The method of claim 10, wherein the tracked duration comprises an amount of time between the message identification and either an identification of an acknowledging message from an assigned node in the data delivery system or a termination of a time-out period during which no acknowledging message is returned by the assigned node.

12. The method of claim 11, wherein the adjusting comprises:

identifying the assigned node as overloaded; and

removing the assigned node from contention for new connections.

13. The method of claim 12, wherein the node-overloaded identification comprises comparing the tracked duration to an historical average for the assigned node.

14. The method of claim 12, wherein the removing comprises transmitting a message to initiate a removal.

15. The method of claim 10, wherein the adjusting of assignment of connection requests is further based upon a bandwidth consumption.

16. The method of claim 15, wherein the adjusting comprises:

identifying an assigned node as overloaded if the tracked duration exceeds a maximum time;

removing the assigned node from contention for new connections if the assigned node is identified as overloaded;

identifying a removed node as available to service new connection requests if a bandwidth consumption for the removed node drops; and

reinstating the removed node to receive new connection requests if the removed node is identified as available to service new connection requests.

17. The method of claim 16, wherein the adjusting further comprises:

identifying the assigned node as overloaded if a number of connections for the assigned node exceeds a maximum connections;

identifying the assigned node as overloaded if the bandwidth consumption for the assigned node exceeds a maximum bandwidth; and

identifying the removed node as available to service new connection requests if the number of connections for the removed node drops below the maximum connections.

18. The method of claim 17, wherein the maximum time, the maximum connections and the maximum bandwidth are each node-specific.

19. The method of claim 10, wherein the data delivery system comprises a web server cluster.

20. Machine-readable instructions, embodied in a machine-readable medium or a propagated signal, for causing a machine to perform operations comprising:

monitoring messages on a network;

identifying a message requesting a connection with a data delivery system having multiple nodes;

initiating a tracking of a duration after the message identification; and

adjusting assignment of connection requests based upon the tracked duration.

21. The instructions of claim 20, wherein the tracked duration comprises an amount of time between the message identification and either an identification of an acknowledging message from an assigned node in the data delivery system or a termination of a time-out period during which no acknowledging message is returned by the assigned node.

22. The instructions of claim 21, wherein the adjusting comprises:

identifying the assigned node as overloaded; and

removing the assigned node from contention for new connections.

23. The instructions of claim 22, wherein the node-overloaded identification comprises comparing the tracked duration to an historical average for the assigned node.

24. The instructions of claim 22, wherein the removing comprises transmitting a message to initiate a removal.

25. A system comprising:

means for accessing messages on a network;

means for storing data from the messages; and

processing means for initiating tracking a duration after a message requesting a connection with a data delivery system having a plurality of nodes, identifying a node from the data delivery system as overloaded based upon the tracked duration, and removing the node from contention for new connections if the node is identified as overloaded.

26. The system of claim 25, wherein the removing comprises transmitting a message to initiate a removal.

27. The system of claim 25, wherein the tracked duration comprises an amount of time between the requesting message and either an acknowledging message from the node or a time-out period during which no acknowledging message is returned by the node.

28. The system of claim 27, wherein the identifying comprises comparing the tracked duration to an historical average for the node.

29. A method of adjusting assignment of connection requests to nodes in a data delivery system, the method comprising:

monitoring messages on a network;

tracking a current number of open connections for each node;

calculating a current bandwidth consumption for each node; and

adjusting assignment of connection requests based upon the tracked connections and the calculated bandwidth consumption.

30. The method of claim 29, wherein the adjusting comprises transmitting a message to initiate an adjustment.