Distributed health check for global server load balancing

• US 20100121932A1
• Filed: 11/27/2002
• Published: 05/13/2010
• Est. Priority Date: 09/26/2000
• Status: Active Grant

• Current Assignee: Avago Technologies International Sales Pte Limited (Broadcom, Inc.)
• Original Assignee: Foundry Networks Incorporated (Broadcom, Inc.)
• Inventors: Joshi, Prajakta S., Cheung, David Chun Ying
• Granted Patent: US 9,130,954 B2
• Time in Patent Office: Days
• US Class Current: 709/207
• 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 67/1012 based on compliance of requirements or conditions with available server resources
• CPC Class Codes: H04L 67/1021 based on client or server locations

Abstract

A global server load-balancing (GSLB) switch serves as a proxy to an authoritative DNS and communicates with numerous site switches that are coupled to host servers serving specific applications. The GSLB switch receives from site switches operational information regarding host servers within the site switches neighborhood. This operational information includes health check information that is remotely obtained in a distributed manner from remote metric agents at the site switches. When a client program requests a resolution of a host name, the GSLB switch, acting as a proxy of an authoritative DNS, returns one or more ordered IP addresses for the host name. The IP addresses are ordered using metrics, including the health check metric that evaluates these IP addresses based on the health check information communicated to the GSLB switch in a distributed manner by the distributed health check site switches. In one instance, the GSLB switch places the address that is deemed “best” at the top of the list.

Claims

1. A system to balance load among host servers, the system comprising:

a load balance switch adapted to be communicatively coupled to a plurality of site switches, the plurality of site switches being remote from the load balance switch, wherein at least one of the site switches is adapted to obtain information indicative of access conditions to its one or more respective host servers and is adapted to provide the obtained information to the load balance switch through a continuous connection between said at least one site switch and said load balance switch and is adapted to further switch traffic to a selected one of said one or more respective host servers; and

a metric collector at the load balance switch to receive the obtained information provided by the at least one site switch, the load balance switch being configured to arrange a list of network addresses based at least in part on the obtained information provided by said at least one site switch.

2. The system of claim 1 wherein the at least one site switch is adapted to provide the obtained information to the load balance switch as part of a message under a communication protocol through the continuous connection and independent of a query from the load balance switch for the obtained information.

3. The system of claim 2 wherein the communication protocol maintains the continuous connection between the at least one site switch and the load balance switch using a keep-alive message type to allow communication of the message having the obtained information via the continuous connection, without having to establish a communication to separately convey information for each port of the one or more host servers.

4. The system of claim 2 wherein the message includes a list of addresses associated with the at least one site switch and an indication of which ports corresponding to addresses in that list are up and are down.

5. The system of claim 4 wherein the message further includes at least one of information indicative of health status of applications hosted on the one or more host servers and configuration changes associated with the at least one site switch.

6. The system of claim 4 wherein the ports corresponding to the list of addresses in the message comprise virtual ports that may be associated with one or more real ports belonging to said one or more host servers that are load balanced by the at least one site switch.

7. The system of claim 1 wherein the at least one site switch obtains the information as part of a distributed health check technique independent of a query for the information by the load balance switch, and wherein:

if there is at least one other of said site switches that is not adapted to support the distributed health check technique and the load balance switch is adapted to support the distributed health check technique, the load balance switch is adapted to query that site switch for information indicative of access conditions; and

if there is at least another of said site switches that is adapted to support the distributed health check technique and the load balance switch is not enabled support the distributed health check technique, said another site switch is adapted to provide information indicative of access conditions to the load balance switch via a centralized health check technique carried out by the load balance switch and is adapted to modify other information related to the information indicative of access conditions into a format that is adapted to be processed by the load balance switch.

8. The system of claim 7 wherein the load balance switch is adapted to enable a flashback metric, which is indicative of a time to respond to a query for the information indicative of access conditions, to arrange said list of network addresses based on the flashback metric and on the information indicative of access conditions, if either the load balance switch or at least some of the plurality of site switches is not enabled to support the distributed health check technique, the flashback metric being disabled otherwise, including if the load balance switch and said at least one site switch support the distributed health check technique.

9. The system of claim 1 wherein an interval of time to provide the obtained information from the at least one site switch to the load balance switch is adapted to be user-specified either globally for all site switches at the load balance switch or individually for each site switch at that respective site switch.

10. The system of claim 1 wherein the at least one site switch includes:

a metric agent to communicate with the metric collector at the load balance switch; and

a health check module in cooperation with the metric agent to obtain the information indicative of access conditions.

11. The system of claim 1 wherein the obtained information includes health check information indicative of health status of ports associated with said one or more host servers for the at least one site switch.

33. The system of claim 1 wherein said list of network addresses is a list of virtual IP addresses, at least one of said virtual IP addresses being configured at each respective site switch and each of said virtual IP addresses corresponding to at least one of said host servers of said respective site switch, wherein said load balance switch is adapted perform said arrangement of said list of virtual IP address to balance traffic between said site switches.

12. A system to balance load in a network, the system comprising:

a site switch adapted to enable distribution of collection of health check information from a remote load balance switch to said site switch, the health check information being indicative of access conditions to components for which said site switch collects health check information and for which said site switch performs switching; and

a metric agent at said site switch to communicate said health check information to the load balance switch using a continuous connection between said site switch and said load balance switch.

13. The system of claim 12 wherein the health check information indicative of access conditions to components includes health check information indicative of access conditions to an application hosted on one or more host servers, each coupled to said site switch.

14. The system of claim 12 wherein the health check information indicative of access conditions to components includes health check information indicative of which ports of one or more host servers coupled to said site switch are up and are down.

15. The system of claim 12 wherein said site switch is adapted to be communicatively coupled to the load balance switch in a manner that the collected health check information of access conditions for all ports associated with said site switch is adapted to be conveyed to the load balance switch as part of a message along with a keep-alive message type that are sent in the continuous connection.

16. The system of claim 12, further comprising a plurality of other site switches that are not adapted support distributed health check information collection, such plurality of other site switches being adapted to be identified by the load balance switch in a manner to allow the load balance switch to process their respective health check information.

17. The system of claim 12, further comprising another load balance switch that is not enabled to support distributed health check information collection by its corresponding site switches, said corresponding site switches being coupled to respond to health check queries from the another load balance switch and to modify other information into a format that is adapted to be processed by said another load balance switch.

18. The system of claim 12 wherein a time interval to send the distributed health check information from said site switch to the load balance switch either is adapted to be specified globally for all site switches at the load balance switch or is adapted to be specified individually for each site switch at each respective site switch.

26. The system of claim 12 wherein a flashback metric representing a time to respond to a query, sent from said load balance switch to said at least site switch, for the health check information is disabled if the collection of the health check information is distributed from said load balance switch to said at least one site switch.

32. The system of claim 12 wherein said metric agent at said site switch is adapted to communicate said health check information to the load balance switch independent of a query for the health check information from the load balance switch.

19. A method of providing load balancing in a network, the method comprising:

establishing a persistent connection between a load balance switch and at least one site switch remote from said load balance switch;

receiving, at the load balance switch through the persistent connection, health check information collected at the at least one site switch and indicative of access conditions to respective host servers for which the at least one site switch performs switching;

arranging at said load balance switch network addresses as an ordered list in accordance with a set of performance metrics, including the health check information collected by and received from the at least one site switch; and

sending from said load balance switch the ordered list of network addresses as a response to a query.

20. The method of claim 19, further comprising disabling a flashback metric, from among the set of performance metrics, indicative of a time to respond to a health check request sent by the load balance switch if said at least one site switch collects said health check information.

21. The method of claim 19 wherein the health check information includes addresses associated with each of said at least one site switch and status indicative of which ports corresponding to the addresses are up and are down, the addresses and status being present in a message sent from each of said at least one site switch, wherein receiving the health check information collected by the at least one site switch through the persistent connection includes:

maintaining the persistent connection between the load balance switch and each of the at least one site switch using a keep-alive message type;

receiving, from the site switches, the message via each persistent connection and independently of a query for said message from said load balance switch; and

receiving, from each of the at least one site switch and independently of a query for said message from said load balance switch, an update to the health check information in an additional message after a specified time interval, and including information indicative of a change in addresses or ports at the each of the at least one site switch, if such a change is present.

22. The method of claim 19, further comprising identifying other site switches that collect health check information in response to requests from the load balance switch.

23. The method of claim 19, further comprising for another load balance switch that is not enabled to support distributed collection of health check information by remote switches:

receiving modified health-check-related information sent by these remote switches to said another load balance switch in a format that is adapted to be processed by said another load balance switch; and

wherein receiving the health-check-related information sent by these remote switches to said another load balance switch includes using a centralized health-check collection technique supported by said another load balance switch.

24. The method of claim 19, further comprising:

specifying, at the load balance switch and globally for all of said at least one site switch, a time interval to provide the health check information to the load balance switch;

or

specifying, individually for each of said at least one site switch, the time interval to be used by that specific site switch to provide the health check information to the load balance switch.

25. The method of claim 19 wherein the received health check information, collected at the at least one site switch, includes information indicative of access to applications hosted at the host servers.

34. The method of claim 19 wherein said list of network addresses is a list of virtual IP addresses, at least one of said virtual IP addresses being configured at said at least one site switch and corresponding to at least one of said host servers of said at least one site switch, wherein said load balance switch is adapted perform said arranging of said list of virtual IP address to balance traffic between said a plurality of said at least one site switch.

27. An article of manufacture, comprising:

a storage medium having instructions stored thereon that are executable by a load balance switch to;

establish a persistent connection from said load balance switch to each of at least some of a plurality of site switches remote from said load balance switch;

process at said load balance switch health check information remotely collected by said at least some of the plurality site switches and indicative of access conditions to respective host servers for which said at least some of the site switches perform switching, the health check information being received by said load balance switch through the persistent connection to each of said at least some of the site switches;

arrange at said load balance switch network addresses as an ordered list in accordance with a set of performance metrics, including the health check information; and

send from said load balance switch the ordered list of network addresses as a response to a query.

28. The article of manufacture of claim 27 wherein the instructions to establish the persistent connection include instructions to use a keep-alive message type to maintain the persistent connection to enable said health check information to be conveyed to said load balance switch independently of a query for said health check information by said load balance switch, without having to establish new connection to separately convey health check information from said at least some of the site switches.

29. The article of manufacture of claim 27 wherein the health check information includes addresses associated with each of said at least some of the site switches and includes indications of which ports associated with each of the addresses are up and are down.

30. The article of manufacture of claim 27 wherein one of the performance metrics includes a flashback metric representing a speed to respond to a request for health check information from the load balance switch, the flashback metric being disabled if the health check information is remotely collected by said at least some of the site switches.

31. The article of manufacture of claim 27 wherein a time interval for said at least some of the site switches to provide the health check information is individually specified for each of said at least some of the site switches.

35. The article of manufacture of claim 27 wherein said list of network addresses is a list of virtual IP addresses, at least one of said virtual IP addresses being configured at each respective site switch and each of said virtual IP addresses corresponding to at least one of said host servers of said respective site switch, wherein said load balance switch is adapted perform said arrangement of said list of virtual IP address to balance traffic between said site switches.