Application delivery controller and global server load balancer

Abstract

Application Delivery Controller (ADC), Global Server Load Balancer (GSLB), and methods for their operation in data networks are disclosed. The methods for load balancing may include receiving a query concerning a host name from a client, determining that there are two or more host servers associated with the host name, measuring various metrics associated with each of the two or more host servers and a local Doman Name Server (DNS), and based at least in part on the measurement, selecting a host server among the two or more host servers. The load balancing may also be based on a measured round trip time.

Description

BACKGROUND

The present disclosure relates generally to data processing, more specifically to Application Delivery Controllers (ADC) and Global Server Load Balancers (GSLB).

Websites, web and mobile applications, cloud computing, and various web and mobile services have been rising in popularity. Some examples of fast growing consumer services include smart phone applications, location based services, navigation services, e-book services, video applications, music applications, Internet television services, and so forth. Subsequently, more and more servers are deployed within data networks including the Internet to accommodate the increasing computing and data storage needs. These servers are typically arranged in data centers or web farms, which may include ADCs, GSLB and/or server load balancers (SLBs).

Conventionally, an ADC is a network device disposed in a datacenter and part of an application delivery network (ADN). The ADC may allow performing common tasks, normally done by web servers, in an effort to remove some load from the web servers. ADCs are typically placed between the firewall/router and the host (web) servers. In addition, conventional ADCs may include various features providing for compression, caching, connection multiplexing, application layer security, and content switching. These features may be combined with basic server load balancing, content manipulation, advanced routing strategies, and highly configurable server health monitoring.

Additionally, ADCs may manage load balancing and delivery of service sessions from client host computers to servers based at least in part on incoming service requests. As more servers are deployed, additional ADC's may be deployed. Similarly, as more servers are pooled together within the data center or spread across multiple data centers to provide scalability, ADCs may become bottlenecks slowing data transmissions between peers on the network.

SUMMARY

This summary is provided to introduce a selection of concepts in a simplified form that are further described in the Detailed Description below. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

According to another aspect of the present disclosure, a GSLB and method of its operation are provided. Specifically, there is provided a method for load balancing between host servers of a data network, the method including receiving a query concerning a host name from a client, determining that there are two or more host servers associated with the host name, measuring round trip times associated with site switches and a Doman Name Server (DNS) assigned to the client, and based at least in part on the measurements, selecting a host server among the two or more host servers. The method may further include providing a network address of the selected host server in the DNS response. The round trip time may include a time for exchange of at least one message from multiple site switches and the Local DNS servers. According to this method, the host server associated with the shortest round trip time is selected from among the two or more host servers.

According to yet another aspect of the present disclosure, another GSLB and method of its operation are provided. In particular, there may be provided a method for load balancing among host servers of a data network. The method may include measuring multiple performance metrics concerning a plurality of switches, each of which may be associated with one or more host servers. The method may further include determining a plurality of network addresses associated with the one or more host servers and storing the multiple performance metrics in association with the plurality of network addresses in a table. The multiple performance metrics may include a plurality of round trip times from a plurality of plurality of switches associated with one or more host servers and a DNS associated with the client. The multiple performance metrics may also include application health metrics, load metrics, proximity metrics, weighted preferences metrics. The method may further include receiving a domain name query from a client or a DNS, retrieving a plurality of network addresses associated with the domain name query from the table, retrieving multiple performance metrics for each network address from the table, randomly selecting one of the network addresses, and calculating a score associated with the multiple performance metrics related to the selected network address.

Furthermore, the method may include determining that the score for the randomly selected network address meets or exceeds a predetermined threshold score and, based on the determination, returning the randomly selected network address to the client or the DNS. If the score for the randomly selected network address does not meet the predetermined threshold score, the method may proceed with removing the randomly network address from the table (although the address need not be removed) and continuing with randomly selecting one of the remaining network addresses from the table to repeat the steps of calculating a score and matching it to the threshold value. If no addresses meet or exceed the threshold than the method may decrease the predetermined threshold and repeat the above steps.

The systems and methods of the present disclosure may be practiced with various electronic devices including, for example, host servers, web farms, switches, routers, client computers such as laptop computers, desktop computers, tablet computers, cellular phones, and other consumer electronic user devices having network connectivity. These and other embodiments are described further below with references to the figures.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments are illustrated by way of example, and not limitation, in the figures of the accompanying drawings, in which like references indicate similar elements.

FIG. 1 is a block diagram of a network suitable for implementing one or more methods of the present disclosure.

FIG. 2 is a flow chart of an exemplary method for operating ADC.

FIG. 3 is another block diagram of an exemplary network suitable for implementing one or more methods of the present disclosure.

FIG. 4 is a flow chart of an exemplary method for load balancing between host servers of a data network.

FIG. 5 is a flow chart of an exemplary method for collecting performance metrics associated with multiple host servers and/or switches.

FIG. 6 is a high-level diagram of exemplary table of aggregated performance metrics associated with multiple host servers and/or switches.

DETAILED DESCRIPTION

In the following description, numerous specific details are set forth in order to provide a thorough understanding of the presented concepts. The presented concepts may be practiced without some or all of these specific details. In other instances, well known process operations have not been described in detail so as to not unnecessarily obscure the described concepts. While some concepts will be described in conjunction with the specific embodiments, it will be understood that these embodiments are not intended to be limiting.

Embodiments disclosed herein may be implemented using a variety of technologies. For example, the methods described herein may be implemented in software executing on a computer system or in hardware utilizing either a combination of microprocessors or other specially designed application-specific integrated circuits (ASICs), programmable logic devices like FPGA's, or various combinations thereof. In particular, the methods described herein may be implemented by a series of computer-executable instructions residing on a storage medium such as a disk drive, or computer-readable medium. It should be noted that methods disclosed herein can be implemented by a computer, e.g., a desktop computer, tablet computer, laptop computer, smartphone and so forth.

The present technology provides various methods for operation of ADCs and GSLBs in data networks such as the Internet including a plurality of switches, routers, virtual switches, web farms, host servers, and other units. The present technology provides enhanced performance of ADC and allows implementing scalable business solutions for any services, applications, clouds and organizations. Furthermore, the present technology provides a scalable, high-performance application networking platform, which can deliver superior reliability and energy efficiency at lower total cost of ownership. ADC can also provide increased infrastructure efficiency, a faster end user experience, comprehensive Layer 4-7 feature set and flexible virtualization technologies such as Virtual Chassis System, multi-tenancy, and more for public, private and hybrid cloud environments. The ADC and GSLB may include software and/or hardware components/platforms that may vary depending on a particular application, performance, infrastructure, network capacity, data traffic parameters, and so forth. Some example topologies for ADC and/or GSLB are described in U.S. utility patent application Ser. No. 13/363,055, filed on Jan. 31, 2012, titled “Virtual application delivery chassis system” (now U.S. Pat. No. 8,266,235), U.S. utility patent application Ser. No. 13/004,861, filed on Jan. 11, 2011, titled “Virtual Application Delivery Chassis System,” U.S. utility patent application Ser. No. 13/154,399, filed on Jun. 6, 2011, titled “Synchronization of configuration file of virtual application distribution chassis,” U.S. utility patent application Ser. No. 12/958,435, filed on Dec. 2, 2010, titled “System and Method for Distributing Application Traffic to Servers Based on Dynamic Service Response Time,” U.S. utility patent application Ser. No. 12/894,142, filed on Sep. 30, 2010, titled “System and method to balance servers based on server load status,” all of which are incorporated herein by reference in their entireties.

Turning now to FIG. 1, a high-level block diagram of a network topology 100 suitable for implementing one or more methods of the present disclosure is shown. The network topology 100 shown by FIG. 1 may include a number of host servers 105, a number of switches 110 combining/coupling the host servers 105 and thus performing Layer 2 aggregation and corresponding switching. The topology 100 may further include an ADC 115 including one (or more) ADC switches 120. The ADC switches may operate in different modes. such as standalone, active/standby mode, Active-Active and others.

Still referring to FIG. 1, the topology 100 may further include a communications network, which may refer to, for example, the Internet, Local Area Network (LAN), Wide Area Network (WAN), Internet, a cellular network, a telephone network, or any other switched network or their combinations. There is also a plurality of clients 130, which may include end user computers, mobile phones, thin clients, and so forth. There are also one or more Local DNS Server which may be associated with one or more clients 130 and/or one or more host servers 105. As shown in FIG. 1, the topology may include a GSLB 135, which may employ one or more of the methods disclosed herein.

Generally speaking, load balancing is a technique that may be used for distributing the workload evenly across clients 130, networks 125, host servers 105, and other networked resources. The load balancing may enhance utilization of resources and enable maximize throughput with minimum response time, hence avoiding overloading of a single server. GSLB may be considered an extension of the load balancing. With this technology, network traffic may be distributed among different web farms, data centers, and host servers 105 located at different geographical locations. This technology may be highly efficient in avoiding local downtimes and downtimes. Furthermore, as will be appreciated by those skilled in the art, GSLB 135 may act as a master to monitor “health” and responsiveness of other sites hosted by the host servers 105. The GSLB may include redirection of service requests to other nearby host servers 105 if one of the host servers 105 does not respond timely. Furthermore, this technique may allow forwarding visitor requests to the host server 105 located most closely geographically to the place from where the request sent. In addition, if a traffic threshold is reached at this host server 105, the service requests may be forwarded to other host server 105 located at a different geographical location.

As will be appreciated by those skilled in the art, these ADC switches 120 may operate in an active mode, backup mode, or some other modes depending on an application. The ADC switches 120 may also provide redundancy protection and failover protection for selected networks or parts of the network 125. The ADC switches 120 may also report their status (i.e., current operating mode) to selected network elements or other switches 110, 120.

FIG. 2 illustrates a block diagram of an exemplary network topology 200 in which various embodiments of the present technology may be practiced. Specifically, FIG. 2 illustrates a specific embodiment of the network topology 100 shown generally in FIG. 1 as well as operations of the GSLB 135.

As shown in FIG. 2, the networked elements may include various arrangements within the network topology and may be located at different geographic locations. Specifically, there may be multiple host server site switches 110 coupling many host servers 105 to the client 130 via the communications network 125. As discussed above, the GSLB 135 may perform load balancing of traffic among the host servers 105. As such, the time for exchanging of a message between the host server site switch 110 and client device 130 may be variable based on at least the capacity of the host servers, the overall traffic load, and the time delay of transmitting a message through the network 125.

In the configuration shown in FIG. 2, the resources or data requested by the client 130, or the resources or data to be to transmitted to the client 130 may be stored in host servers 105. As illustrated in the figure, Host Server 1, Host Server 2, and corresponding Switch 1 may be located in Seattle, Wash., while Host Server 3, Host Server 4, and corresponding Switch 2 may be located in New York City, N.Y. By way of example, the client 130 may be located in Pennsylvania, Pa. and the corresponding local DNS 140 may be located in Los Angeles, Calif. In such configurations, the GSLB 135 may perform corresponding measurements to determine which host server 105 should be used for transfer data to the client 130. The measurements, in some embodiments, include round trip times of exchange of a message between the switches 110 and the local DNS 140. In this configuration, the round trip time may be a function of network or switch capacity, the overall traffic load, time delay of transmitting a message through the network, and so forth. Thus, as will be appreciated by those skilled in the art, the time for exchanging a message between Switch 1 and the local DNS 140 is substantially a measurement of the time to send a message between Seattle and Los Angeles. The local DNS 140 in Los Angeles is much closer geographically to Switch 1 in Seattle, than the client 130 is in Philadelphia. Thus, the time for a message to travel from Switch 1 to the local DNS 140 and back should be less than the time for a message to travel from Switch 1 to the client 130 and back. In this regard, the GSLB 135 may select “Switch 1” 110 to make data transmission to the client 130, and not the host servers 105 or Switch 2, which located closer to the client 130. These principles are further described in the following figure.

FIG. 3 illustrates a flow chart of an example method 300 for load balancing between host servers 105 of a data network 125. The method 300 may be practiced by the GSLB 135 as described above with references to FIGS. 1 and 2.

The method 300 may commence in operation 310 with the GSLB 135 receiving a query concerning a host name from a client 130. This query, in certain embodiments, may be generated by the ADC 115 so as to initiate the load balancing procedure.

In operation 320, the GSLB 135 may determine that there are two or more of host servers 105 associated with the host name. If there is just one host server 105, the load balancing procedure is not performed. Otherwise, the method 300 may proceed to operation 330, so that the GSLB 135 may measure round trip times (or other similar metrics) associated with each of the two or more host servers 105 and the LDNS 140. More specifically, round trip times may be measured for exchange of at least one message from each of the two or more host servers 105 and the LDNS 140. In certain embodiments, the round trip time may also include a time for exchange of at least one message from one or more switches 110 associated with the two or more host servers 105 and the LDNS 140.

At operation 340, based at least in part on the measurement, the GSLB 135 may select the host server 105, from the two or more host servers 105, which is associated with the shortest round trip time measured. At operation 350, the GSLB 135 may provide a network address of the selected host server 105 to the LDNS 140 or the client 130 so that data may be transmitted from the selected host server 105 to the client 130.

In certain additional embodiments, the measured round trip times may be stored in the GSLB 135 and then retrieved upon further request. This procedure may eliminate frequent redundant operations and thus may save some computational and networked resources.

According to one or more embodiments of the present disclosure, the global load balancing process may rely not only on real time measurement of various parameters, such as described above, but also on various measurements performed in advance. FIG. 4 shows a flow chart of an example method 400 for collecting performance metrics associated with certain host servers 105 and/or switches 110. The method 400 may be practiced by the ADC 115 and/or the GSLB 135 and/or similar electronic devices as discussed above.

The method 400 may include operation 410, at which the GSLB 135 (or similar device) may measure multiple performance metrics regarding a plurality of switches 110, and each switch 110 may be associated with one or more host servers 105. The performance metrics may include a plurality of round trip times from the plurality of switches 110 associated with one or more host servers 105 and a DNS 140 associated with the client 130. In certain embodiments, the performance metrics may include one or more of the following: application health metrics, load metrics, and proximity metrics.

Turning again to FIG. 4, in operation 420 the GSLB 135 may also determine a plurality of network addresses associated with the one or more host servers 105, with respect to which the measurements have been performed. In operation 530, the GSLB 135 may store the multiple performance metrics in association with the plurality of network addresses in a table. An example of such a table is shown in FIG. 5.

In particular, FIG. 5 shows a high level diagram of example table 500 of aggregated performance metrics associated with certain host servers 105 and/or switches 110. The table cross-references measured metrics with networked addresses. The metrics may be binary or measured values may be provided. The table 500 may be used in the process described below with reference to FIG. 6.

FIG. 6 illustrates a flow chart of an example method 600 for load balancing among host servers 105 of a data network 125. The method 600 may be practiced by the GSLB 135 as described above with reference to FIGS. 1 and 2.

Method 600 may commence in operation 610 with the GSLB 135 receiving a domain name query from a client 130 or a DNS 140. In operation 620, the GSLB 135 may further retrieve a plurality of network addresses associated with the domain name query as well as multiple performance metrics for each network address. The retrieving may be performed from the table 600 as shown in FIG. 6.

In operation 630, the GSLB 135 may randomly select one of the network addresses from the table 600 and calculate a score associated with the multiple performance metrics related to the selected network address. In operation 640, the GSLB 135 may further compare the score for the randomly selected network address to a predetermined threshold value. If it is determined that the score meets or exceeds the threshold value, the method 600 may proceed to operation 650, where the GSLB 135 may transmit the randomly selected network address to the client 130 or the DNS 140. Otherwise, if it is determined that the score does not meet or exceed the threshold value, the method 600 may proceed to operation 670, where the GSLB 135 may remove the randomly selected network address from the table 600 and then, in operation 660, check if there are any other network addresses remaining in the table 600. If there are remaining network addresses in the table 600, the method 600 may proceed to operation 630 and repeat random selection of another network address, its corresponding score, and continue with the determination at operation 640. Otherwise, if there are no remaining network addresses in the table 600, the method proceeds to operation 680, when the GSLB 135 may decrease the threshold value a certain amount and then the method 600 may return to operations 630 and 640 until a network address generating a qualifying score is selected.

Accordingly, the present technology for global load balancing sets a threshold score and evaluates a randomly accessed network address from a plurality of network addresses against the threshold score until an address is found that meets or exceeds the threshold score. There is no comparison of network addresses against each other, and thus there is no ranking or ordering of network addresses and no generation of an ordered list of network addresses. This approach significantly simplifies and facilitates the load balancing process.

It should be noted that the systems and methods herein may return multiple network addresses in response to a given domain name query. Each of the returned addresses will be randomly selected and non-ordered, but each returned address will meet the threshold limit. If it is desirable in a given application that multiple network addresses be returned, the qualification procedure described above may be performed multiple times. Alternatively, a randomly selected network address associated with a given domain name query may be stored for future reference. Should the same domain name query then be re-submitted, so that the query is a recognized query, the systems and methods described herein may return one or more stored addresses in addition to one or more newly selected network addresses.

Thus, methods and systems for operation of the ADC and GSLB have been described. Although embodiments have been described with reference to specific example embodiments, it will be evident that various modifications and changes can be made to these example embodiments without departing from the broader spirit and scope of the present application. Accordingly, the specification and drawings are to be regarded in an illustrative rather than a restrictive sense.

Claims

1. A method for load balancing among host servers of a data network, the method comprising: receiving, by a Global Server Load Balancer (GSLB), a domain name query from a client; retrieving, by the GSLB, a plurality of network addresses associated with the domain name query, the plurality of network addresses being associated with a plurality of host servers; upon the retrieving the plurality of network addresses;determining, by the GSLB, at least a round trip time between a Local Domain Name Server (LDNS) associated with the client and each of the plurality of host servers associated with the plurality of network addresses, the determining including exchanging at least one message between the LDNS and each of the plurality of host servers; anddetermining, by the GSLB, geographic distances between the LDNS and each of the plurality of host servers;based on the determining of both the round trip time and the geographic distances, storing, by the GSLB, the round trip time for each of the plurality of network addresses and the geographic distances to a metrics table, the metrics table storing multiple performance metrics for each of the plurality of network addresses;randomly selecting, by the GSLB, a network address of the plurality of network addresses, the randomly selected network address having performance metrics selected from the multiple performance metrics;scoring the performance metrics of the randomly selected network address to obtain an aggregated metrics score of the randomly selected network address;determining, by the GSLB, whether the aggregated metrics score of the randomly selected network address meets a predetermined threshold; andbased on the determination that the aggregated metrics score meets the predetermined threshold, returning, by the GSLB, the randomly selected network address.

2. The method of claim 1, wherein the determining that the aggregated metrics score of the randomly selected network address meets the predetermined threshold includes: determining that the aggregated metrics score is higher than the predetermined threshold.

3. The method of claim 1, wherein the GSLB is configured to randomly select a further network address of the plurality of network addresses based on the determination that the aggregated metrics score of the randomly selected network address does not meet the predetermined threshold.

4. The method of claim 2, further comprising: determining, by the GSLB, aggregated metrics scores for all of the plurality of network addresses;determining, by the GSLB, that the aggregated metrics scores for all of the plurality of network addresses do not meet the predetermined threshold; and

5. The method of claim 2, wherein the domain name query is received from a client or a Domain Name Server (DNS).

6. The method of claim 5, further comprising: determining, by the GSLB, a plurality of performance metrics associated with a plurality of switches, wherein each of the plurality of switches is associated with one or more host servers;determining, by the GSLB, multiple network addresses associated with the one or more host servers; andstoring, by the GSLB, the plurality of performance metrics in association with the multiple network addresses in a table.

7. The method of claim 5, further comprising: randomly selecting, by the GSLB, a further network address of the plurality of network addresses, the randomly selected further network address having further performance metrics selected from the multiple performance metrics; scoring the further performance metrics of the randomly selected further network address to obtain a further aggregated metrics score of the randomly selected further network address;based on the multiple performance metrics, determining, by the GSLB, that the further aggregated metrics score of the randomly selected further network address meets the predetermined threshold; andbased on the determination, returning, by the GSLB, the randomly selected further network address so that at least two randomly selected network addresses are returned to the client or the DNS, the at least two randomly selected network addresses including the randomly selected network address and the randomly selected further network address.

8. The method of claim 5, further comprising: receiving a recognized domain name query;randomly selecting, by the GSLB, at least one new randomly selected network address, the least one new randomly selected network address having further performance metrics selected from the multiple performance metrics;scoring the further performance metrics of the least one new randomly selected network address to obtain a further aggregated metrics score of the least one new randomly selected network address;based on the multiple performance metrics, determining, by the GSLB, that the further aggregated metrics score of the least one new randomly selected further network address meets the predetermined threshold; andbased on the determination, returning both the at least one new randomly selected network address and the randomly selected network address to the client or the DNS.

9. The method of claim 6, wherein the multiple performance metrics includes a plurality of round trip times from the plurality of switches associated with the one or more host servers and the LDNS associated with the client.

10. The method of claim 1, wherein the multiple performance metrics include one or more of the following: application health metrics, proximity metrics, and weighted preferences metrics.

11. A method for load balancing among host servers of a data network, the method comprising: receiving, by a global server load balancer (GSLB), a query concerning a host name from a client;determining, by the GSLB, that there are two or more site switches and two or more host servers associated with the host name;determining, by the GSLB, a round trip time between a Local Domain Name Server (LDNS) assigned to the client and each of the two or more site switches, the determining including exchanging at least one message between the LDNS and each of the two or more site switches and determining, by the GSLB, geographic distances between the LDNS and each of the two or more site switches;based on the determining of both the round trip time and the geographic distances, storing, by the GSLB, the round trip time for each of the two or more site switches and the geographic distances to a metrics table, the metrics table storing multiple performance metrics for each of the two or more site switches;randomly selecting, by the GSLB, a host server from the two or more host servers, the randomly selected host server being associated with a site switch of the two or more site switches;scoring, by the GSLB, the multiple performance metrics of the site switch to obtain an aggregated metrics score of the site switch;determining, by the GSLB, that the aggregated metrics score of the site switch meets a predetermined threshold; andbased on the determination, returning, by the GSLB, a network address of the randomly selected host server associated with the site switch.

12. The method of claim 11, further comprising providing the network address of the randomly selected host server to the LDNS.

13. The method of claim 11, wherein each of the round trip times is a time for an exchange of at least one message from each of the two or more site switches associated with the two or more host servers and the LDNS.

14. The method of claim 11, wherein each of the round trip times is a time for an exchange of at least one message from each of the two or more host servers and the LDNS.

15. The method of claim 11, further comprising transmitting data from the randomly selected host server to the client.

16. The method of claim 11, further comprising: storing the round trip times associated with the two or more host servers in the GSLB; retrieving, by the GSLB, at least one of the round trip times in response to a further query.

17. The method of claim 11, wherein the query concerning the host name is generated by an Application Delivery Controller (ADC) associated with the one or more host servers.

18. The method of claim 11, wherein the randomly selected host server is associated with a shortest round trip time of the round trip times.

19. The method of claim 11, further comprising: randomly selecting, by the GSLB, a further host server from the two or more host servers, the randomly selected further host server being associated with a further site switch of the two or more site switches;scoring the multiple performance metrics of the further site switch to obtain a further aggregated metrics score of the further site switch;determining, by the GSLB, that the further aggregated metrics score of the further site switch meets the predetermined threshold; andbased on the determination, returning, by the GSLB, the further network address of the randomly selected further host server so that at least two randomly selected network addresses are returned to the client or the LDNS, the at least two randomly selected network addresses including the randomly selected network address and the randomly selected further network address.

20. The method of claim 11, further comprising: receiving a recognized query concerning the host name; randomly selecting, by the GSLB, at least one new randomly selected host server, the at least one new randomly selected host server being associated with a further site switch of the two or more site switches;scoring the performance metrics of the further site switch to obtain a further aggregated metrics score of the further site switch;determining, by the GSLB, that the further aggregated metrics score of the further site switch meets the predetermined threshold; andbased on the determination, returning both at least one new network address of the at least one new randomly selected host server and the network address of the randomly selected host server to the client or the LDNS.