Global server load balancing

• US 7,254,626 B1
• Filed: 07/25/2002
• Issued: 08/07/2007
• Est. Priority Date: 09/26/2000
• Status: Expired due to Fees

• Current Assignee: Avago Technologies International Sales Pte Limited (Broadcom, Inc.)
• Original Assignee: Foundry Networks Incorporated (Broadcom, Inc.)
• Inventors: Hsu, Ivy Pei-Shan, Jalan, Rajkumar, Cheung, David Chun Ying, Kommula, Sunanda Lakshmi
• Primary Examiner(s): Etienne; Ario
• Assistant Examiner(s): EL CHANTI, HUSSEIN A
• Application Number:
  US10/206,580
• Time in Patent Office: 1,839 Days
• Field of Search: 709223-225, 709/226, 709/241, 709/201, 709/203, 709/213, 709/217, 709/219, 709/238
• US Class Current: 709/223
• 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. 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 that include the information collected from the site switches. In one instance, the GSLB switch places the address that is deemed “best” at the top of the list.

Claims

1. A method of providing load balancing among host servers using a load balance switch and a plurality of site switches that each couple at least one of the host servers to a network, the method comprising:

collecting, at the site switches, a site-specific metrics indicative of access conditions to the at least one of the host servers coupled to the respective site switch;

receiving, at the load balance switch, the site-specific metrics collected at the site switches for a set of performance metrics,

wherein the set of performance metrics includes a connection-load metric that represents a measure of new connections per second at each respective site switch, the connection-load metric being determined by;

setting a connection load limit;

determining new connections at site switches using a sampling rate and a sampling interval;

assigning a weight to each interval; and

calculating an average load for each of the site switches using the weight assigned to each interval and the determined new connections at each interval;

the method further comprising;

whenever network addresses are provided in response to a query, arranging the network addresses as an ordered list in accordance with the set of performance metrics; and

sending the ordered list of network addresses as a response to the query.

2. The method of claim 1, further comprising:

accepting, for the ordered list of network addresses, site switches that have a calculated average load less than the connection load limit; and

rejecting site switches that have a calculated average load greater than the connection load limit.

3. The method of claim 1 wherein the set of performance metrics includes a geographic metric associated with geographical locations of the host servers, the method further comprising applying the connection load metric after applying the geographic metric to determine the ordered list of network addresses.

4. The method of claim 1, further comprising:

collecting, at the load balance switch, another set of performance metrics; and

including the another set of performance metrics with the set of performance metrics to determine the ordered list of network addresses.

5. The method of claim 4 wherein the another set of performance metrics includes a health check metric associated with a health check sent from the load balance switch to each of the site switches.

6. The method of claim 1 wherein assigning the weight to each interval includes assigning a weight having a value of 1 to at least some of said intervals.

7. The method of claim 1, further comprising nullifying a contribution of a particular interval by assigning a weight having a value of 0 to said particular interval.

8. A method of providing load balancing among host servers using a load balance switch and plurality of site switches that each couple at least one of the host servers to a network, the method comprising:

receiving, from the site switches, site-specific metrics for a set of performance metrics indicative of access conditions to the at least one of the host servers coupled to the respective site switch, wherein the set of performance metrics includes a connection-load metric, said connection-load metric representing a measure of new connections per second at each respective site switch, said connection-load metric being calculated by;

sampling new connections at the site switches based on a sampling rate and a sampling interval;

assigning a weight to each interval;

calculating an average load of new connections for each of the site switches by using the weight assigned to each interval and the new connections of each interval;

the method further including;

ordering, based on the calculated average loads of each of the site switches, a list of network addresses that are responsive to a query, said ordering including;

accepting site switches that have a calculated average load less than a connection load limit; and

rejecting site switches that have a calculated average load greater than the connection load limit.

9. The method of claim 8 wherein ordering the list of network addresses includes arranging the network addresses based at least in part on another performance metric, associated with geographical locations of the host servers, wherein ordering the list of network addresses based on the calculated average load is performed after applying the performance metric associated with geographical locations of the host servers.

10. The method of claim 8 wherein assigning the weight to each interval includes assigning a weight having a value of 1 to at least some of said intervals.

11. The method of claim 8, further comprising nullifying a contribution of a particular interval by assigning a weight having a value of 0 to said particular interval.

12. An article of manufacture usable in a system that provides load balancing among host servers using a load balance switch and a plurality of site switches that each couple at least one of the host servers to a network, the article of manufacture comprising:

a machine-readable medium having instructions stored thereon to;

request, from the site switches, a site-specific metric indicative of new connections per unit of time to the site switches;

whenever network addresses are provided in response to a query, arrange the network addresses as an ordered list based at least in part on the site-specific metric received from the site switches; and

send the ordered list of network addresses as a response to the query, wherein the machine-readable medium is located at the load balance switch and further includes instructions stored thereon to;

set a connection load limit;

request new connection data from the site switches, the connection data being obtained using a sampling rate and a sampling interval;

assign a weight to each interval;

calculate an average load for each of the site switches using the weight assigned to each interval and the new connection data for each interval;

accept, for the ordered list of network addresses, site switches that have a calculated average load less than the connection load limit; and

reject site switches that have a calculated average load greater than the connection load limit.

13. The article of manufacture of claim 12 wherein the instructions to arrange the network addresses as the ordered list include instructions to arrange the network addresses based at least in part on another site-specific metric, associated with geographical locations of the host servers, the machine-readable medium further including instructions stored thereon to apply the site-specific metric indicative of new connections per unit of time after applying the site-specific metric associated with geographical locations of the host servers to obtain the ordered list.

14. The article of manufacture of claim 12 wherein the instructions to assign the weight to each interval includes instructions to assign a weight having a value of 1 to at least some of said intervals.

15. The article of manufacture of claim 12 wherein the machine-readable further includes instructions stored thereon to nullify a contribution of a particular interval by using a weight having a value of 0 for said particular interval.

16. A method of providing load balancing among host servers using a load balance switch and plurality of site switches that each couple at least one of the host servers to a network, the method comprising:

determining a number of new connections per unit time for at least some of the site switches using a sampling rate and a sampling interval; and

arranging a list of network addresses based at least in part on the determined number of new connections per unit time, in a manner that site switches having a lower number of new connections per unit time relative to other site switches are preferentially selected to receive network traffic, the method further comprising;

assigning a weight to at least one of the sampling intervals;

calculating an average load for each site switch using;

a) the weight assigned to the at least one sampling interval, and b) determined new connections corresponding to the at least one sampling interval;

comparing the calculated average load for each site switch with a connection load limit;

accepting, for the list of network addresses, site switches that have a calculated average load less than the connection load limit; and

rejecting site switches that have a calculated average load greater than the connection load limit.

17. The method of claim 16 wherein assigning the weight to at least one of the intervals includes assigning a weight having a value of 1 to said at least one of the intervals.

18. The method of claim 16, further comprising nullifying a contribution of a particular interval by assigning a weight having a value of 0 to said particular interval.

19. An article of manufacture usable to provide load balancing in a network including host servers, the article of manufacture comprising:

a machine-readable storage medium storing instructions to;

store a connection load limit;

calculate, or store a remotely calculated, average load for each of a plurality of site switches, each respective calculated average load reflecting a number of new connections per second to each of the plurality of site switches, each of the site switches for coupling at least one of said host servers to the network, the new connections to the site switches being determined using a sampling rate and a sampling interval, the calculated average load of each respective site switch being calculated by using a weight assigned to each interval and the new connections for each of the intervals;

output a list of network addresses based on the calculated average load of each of said plurality of site switches, including;

accept, for the list of network addresses, any of said site switches that has a calculated average load less than the connection load limit; and

reject site switches that have a calculated average load greater than the connection load limit.

20. The article of manufacture of claim 19 wherein the weight assigned to each interval includes:

a weight, for at least one of the sampling intervals, with a value of 1; and

a weight, for at least another one of the sampling intervals, with a value of 0 to nullify a contribution of said another one of the sampling intervals to the calculated average load.

21. An apparatus to balance load among host servers of a network, the apparatus comprising:

a network device adapted to communicate with a plurality of site switches, each site switch for coupling at least one of said host servers to the network, the network device being configurable to receive from the site switches site-specific metrics for a set of performance metrics that are indicative of access conditions to the respective host servers, wherein the set of performance metrics includes a connection-load metric representing a measure of new connections per second at the respective site switches based on a sampling rate and a sampling interval, the connection-load metric being determinable by the network device which is configurable to calculate an average load of new connections for each of the site switches using a weight assigned to each interval and the new connections of each interval, and to order, based on the calculated average loads of each of the site switches, a list of network addresses that are responsive to a query,

wherein the network device includes a metric collector, and a controller configurable to (a) set a connection load limit, (b) via the metric collector, obtain the new connections per sampling interval from the site switches, and (c) for said ordering, accept site switches that have a calculated average load less than the connection load limit and reject site switches that have a calculated average load greater than the connection load limit.

22. The apparatus of claim 21 wherein the weight has a value of 1.

23. The apparatus of claim 21 wherein the network device is further configurable to arrange the list of network addresses based at least in part on a performance metric corresponding to geographic locations of the host servers, and to evaluate the new connections at the site switches subsequent to evaluation of the performance metric corresponding to the geographic locations of the host servers.

24. The apparatus of claim 21 wherein the set of performance metrics are usable to base arrangement of the list of network addresses, wherein the network device is configurable to change an order of application of the performance metrics to obtain the list of network addresses.

25. The apparatus of claim 24 wherein a default configuration of the network device excludes the new connections at the site switches during application of the performance metrics to obtain the list of network addresses.