SAG clusters

An SAG cluster is a group of SAGs that process outbound request messages that originate from a common source, and so share a common workload. Each SAG cluster is an FTM SWIFT service; that is, it is a message-processing entity whose resource definitions (such as those for queues and security profiles) are managed by FTM SWIFT. A cluster receives business messages and routes them to a network. As with other services, each SAG cluster has its own set of interface and input queues.

Although the concept of an SAG cluster becomes interesting only when you have more than one SAG, even a lone SAG must belong to a cluster. This is because each SAG requires input queues, and these queues always apply to an entire SAG cluster, even if this cluster comprises only a single SAG.

SAG clusters are not defined as FTM SWIFT objects; that is, you do not create a cluster by issuing a "create cluster" command and then specifying the name or other properties of the cluster. Instead, you create a cluster indirectly by defining queue managers and queues in such a way that several SAGs retrieve requests from a common source.

Using SAG clusters composed of more than one SAG:
  • Improves availability, because if one component fails, others assume its workload
  • Increases throughput, because the SAGs in a cluster automatically balance their workloads
  • Makes your configuration scalable, because you can vary the number of SAGs to suit your installations workload

However, the speed and network capacity must be balanced for each SAG in a cluster; otherwise, bottlenecks might result. For example, if one SAG in a cluster runs at high speed but has a low network capacity, it will retrieve a large number of messages from the shared input queue, but will process these messages very slowly. This SAG will develop a backlog, while the others remain idle.

Each cluster can be used by any number of OUs, but must be set up for use by at least one. This follows from the fact that each cluster must have at least one interface queue, and each interface queue corresponds to exactly one OU.

Although all FTM SWIFT servers must run on the same operating system, the queue managers of the SAG input queues can run on any operating system. For example, one such queue manager can run on AIX®, and another on RHEL x86.