Understanding server-level and connection-level fault tolerance

By default, a Unified Broker instance relies on a single controlling NameServer to resolve client connection requests and a single Unified Broker instance to provide services to the client. You can configure the controlling NameServer so that multiple NameServer instances are available to resolve any client connection request, thus providing connection-level fault tolerance. If your product supports load balancing, you can also configure a single NameServer to resolve each connection request using multiple Unified Broker instances that support the same Application Service, thus providing server-level fault tolerance. The following figure shows the relationship between these configuration options.

These two levels of fault tolerance operate as follows:

Server-level fault tolerance — Allows multiple Unified Broker instances to register with a NameServer for the same Application Service. A client requesting a connection is connected to one of several registered Unified Broker instances that the NameServer determines are available to provide the specified Application Service. If appropriate weight factors are specified, the NameServer also balances connection load among the several broker instances. For more information on load balancing, see Understandingload balancing.

Connection-level fault tolerance — Allows you to configure a collection of NameServers that work together to resolve a client connection request. You can use two different techniques, individually or together, to implement a fault-tolerant NameServer collection. This section describes these techniques and assumes that you are familiar with the documentation on using the NameServer with your Unified Broker product.

You can apply server-level and connection-level fault tolerance individually or together to achieve the level of fault tolerance that your application requires.