The answer lies in what is an acceptable level of latency for your customers. A SOA management platform can help you to keep latency within the levels specified by your Service Level Agreements (SLA) that you have with each given customer. It can do this by monitoring and measuring your service request traffic and alerting you to violations. In addition, when a SLA violation occurs, it can help you to perform root cause analysis by reporting timings on individual service calls and specific WSDL operations. In the case of the Actional products, this alerting and reporting can be done within the context of an individual business process, and service requests may also be dynamically routed and prioritized based on higher valued customers. Depending on the architecture of the SOA management platform, the additional latency introduced may be measured in microseconds rather than milliseconds.
If you are using an ESB, and it is built in such a way as to allow the selective control over which parts of the SOA infrastructure can be scaled and load-balanced across multiple machines, that can actually help you to achieve optimal throughput with the lowest latency. In addition asynchronous processing can help smooth out surges and spikes in service request traffic that is otherwise inherent in tightly-coupled synchronous systems. An ESB that provides continuous availability can also ensure consistent quality of service in the face of hardware failures or network disruptions.
This was first published in June 2006