When you integrate a Global Traffic Manager system into your network,
one of its primary responsibilities is to load balance incoming connection requests to the virtual server resource that best fits the configuration parameters you defined. However, load balancing is only one part of managing connections to your network resources. Additional issues that you must consider include:
| || |Resource health
Resource health refers to the ability of a given resource to handle incoming connection requests. For example, the Configuration utility uses a green circle to identify a resource, such as a wide IP, that has available pools and virtual servers, while a pool that is down appears as a red diamond. These visual clues can help you identify connection issues quickly and efficiently.
| || |Resource availability
Resource availability refers to the settings within the Configuration utility that you use to control when a resource is available for connection request. For example, you can establish limit settings, which instruct Global Traffic Manager to consider a resource as unavailable when a statistical threshold (such as CPU usage) is reached.
| || |Restoring availability
When a resource goes offline, Global Traffic Manager immediately sends incoming connection requests to the next applicable resource. When you bring that resource online again, you can control how to restore its availability to Global Traffic Manager, ensuring that connections are sent to the resource only when it is fully ready to receive them.
| || |Persisting connections
Certain interactions with your network require that a given user access the same virtual server resource until their connection is completed. An example of this situation is an online store, in which you want the user to access the same virtual server for their shopping cart until they place their order. With Global Traffic Manager, you can configure your load balancing operations to take persistent connections into account.
| || |Selecting a last resort pool
Global Traffic Manager includes the ability to create a last resort pool. A last resort pool is a collection of virtual servers that are not used during normal load balancing operations. Instead, these virtual servers are held in reserve unless all other pools for a given wide IP become unavailable.
In addition, it is important to understand what happens when Global Traffic
Manager cannot find an available resource with which to respond to a connection request.
In Global Traffic Manager, resource health
refers to the ability of a given resource to handle incoming connection requests. Global Traffic Manager determines this health through the use of limit settings, monitors, and dependencies on other network resources.
The health of a resource is indicated by a status code in the Configuration
utility. A status code
is a visual representation of the availability of a given resource. Global Traffic Manager displays these status codes in the main screens for a given resource. The types of status codes available for a resource are:
| || |Blue
A blue status code indicates that the resource has not been checked. This status often appears when you first add a resource into the Configuration utility.
| || |Green
A green status code indicates that the resource is available and operational. Global Traffic Manager uses this resource to manage traffic as appropriate.
| || |Red
A red status code indicates that the resource did not respond as expected to a monitor. Global Traffic Manager uses this resource only when two conditions are met:
| || |Yellow
A yellow status code indicates that the resource is operational, but has exceeded one of its established bandwidth thresholds. Global Traffic Manager uses a resource that has a yellow status code only if no other resource is available.
| || |Black
A black status code indicates that the resource has been manually disabled and is no longer available for load balancing operations.
As the preceding list illustrates, the health of a resource does not necessarily
impact the availability of that resource. For example, Global Traffic Manager can select a virtual server that has a red status code.
To load balance effectively, Global Traffic Manager must determine
whether the appropriate resources are available. In the context of the
Global Traffic Manager, availability
means that the resource meets one or more sets of pre-defined requirements. These requirements can be a set of statistical thresholds, a dependency on another resource, or set of values returned by a monitoring agent. If a resource fails to meet one or more of these requirements, Global Traffic Manager considers it unavailable
and attempts to select the next resource based on the load balancing methodology you defined.
One of the methods for determining the availability of a resource is to
establish limit settings. A limit setting
is a threshold for a particular statistic associated with a system.
Another method for determining the availability of a given resource is
through the use of monitors. A monitor
is a software utility that specializes in a specific metric of a Global Traffic Manager resource. You can customize monitors to be as specific or as general as needed.
To illustrate the use of monitors to determine the availability of a resource,
consider the fictional company SiteRequest. One of the servers at SiteRequests Paris data center, serverWeb1
, contains the main web site content for the wide IP, www.siterequest.com
. To ensure that this server is available, SiteRequest configures an HTTP monitor within Global Traffic Manager and assigns it to serverWeb1
. This monitor periodically accesses the server to verify that the main index.html
page is available. If the monitor cannot access the page, it notifies Global Traffic Manager, which then considers the server unavailable until the monitor is successful.
Monitors provide a robust, customizable means of determining the
availability of a given resource with Global Traffic Manager. The following procedure describes how to control the impact that a set of monitors has on the availability of a resource.
You can also assign monitors to a specific server. In most cases, when you
assign a monitor to a server, that monitor checks all virtual servers associated with that server.
An exception to this guideline is the SNMP monitor. If you assign an SNMP
monitor to a Cisco®
, Extreme Networks®
, or Radware server, that monitor obtains information on the virtual servers associated with that server. If you assign the SNMP monitor to any other server type, that monitor obtains data on the server itself.
In cases where you assign a monitor to a virtual server both directly and to
its parent server, the availability information acquired from the monitor directly assigned to the virtual server takes precedence over any other data.
Within Global Traffic Manager, you can configure a virtual server to be
dependent on one or more virtual servers. In such a configuration, the virtual server is available only if all of the resources in its Dependency List
are available as well.
For an example of virtual server dependencies, consider the fictional
company SiteRequest. One of the servers, serverMain
, at the Tokyo data center has two virtual servers: vsContact
, which points to the contacts page of SiteRequests web site, and vsMail
, which points to their mail system. The vsContact
virtual server has vsMail
added in its Dependency List
. As a result, Global Traffic Manager considers the vsContact
virtual server available only if the vsMail
virtual server is also available.
You can set dependencies for a virtual server at any time. When you
configure the Dependency List
option for a virtual server, Global Traffic Manager checks each virtual server in the order in which you added it to the Configuration utility. You can change this order at any time.
When a network resource, such as a virtual server, goes offline, Global
Traffic Manager considers that resource to be unavailable and proceeds to send name resolution requests to other resources based on the configured load balancing mode. By default, Global Traffic Manager resumes sending requests to an offline resource as soon as that the resource becomes available again, provided that the resource meets the appropriate load balancing requirements.
Under certain circumstances, you might not want Global Traffic Manager to
resume connections to a resource immediately. For example, a server for the fictional company, SiteRequest, goes offline. Global Traffic Manager detects that the virtual servers associated with this server are unavailable, and proceeds to send name resolution requests to other virtual servers as appropriate. When the server is online again, it must still run several synchronization processes before it is fully ready to handle name resolution requests. However, Global Traffic Manager might detect that the server is available before these processes are complete, and send requests to the server before that server can handle them.
To avoid this possibility, you can configure pools to use the manual resume
feature. The manual resume
feature ensures that Global Traffic Manager does not load balance requests to a virtual server within a pool until you manually re-enable it.
Most load balancing modes divide name resolution requests among
available pools or virtual servers. Each time Global Traffic Manager receives a request, it sends that request to the most appropriate resource based on the configuration of your network. For example, when a user visits a web site, it results in multiple name resolution requests as that user moves from page to page. Depending on the load balancing mode selected, the system sends each request to a completely different server, virtual server, or data center.
In certain circumstances, you might want to ensure that a user remains with
a given set of resources throughout the session. For example, a user attempting to conduct a transaction through an online bank needs to remain with the same set of resources to ensure the transaction is completed successfully.
To ensure that users stay with a specific set of resources, Global Traffic
Manager includes a persistence option. The persistence option instructs the system to send a user to the same set of resources until a specified period of time has elapsed.
If you elect to use persistent connections with a load balancing mode, you
must decide how to handle connection requests when you need to take a specific pool of virtual servers offline. By default, Global Traffic Manager immediately sends connection requests to other pools when you take that pool offline, even if persistent connections are enabled. In some situations, this behavior might not be desirable. For example, consider an online store. You might need to take a pool of virtual servers for this store offline; however, you do not want to interrupt shoppers currently purchasing any products. In this situation, you want to drain persistent requests.
refers to allowing existing sessions to continue accessing a specific set of resources while disallowing new connections. In Global Traffic Manager, you configure this capability through the Drain Persistent Requests
option. This option applies only when you manually disable the pool. It does not apply when the pool goes offline for any other reason.
When Global Traffic Manager load balances name resolution requests, it
considers any pool associated with a given wide IP as a potential resource. You can, however, modify this behavior by creating a last resort pool. A last resort pool
is a pool of virtual servers to which the system sends connection requests in the event that all other pools are unavailable.
It is important to remember that any pool you assign as the last resort pool is
not a part of the normal load balancing operations of Global Traffic Manager. Instead, this pool is kept in reserve. The system uses the resources included in this pool only if no other resources are available to handle the name resolution request.