Trading Networks and Integration Servers


I am new to Web Methods and am trying to architect out how we will be implementing. I am considering using Trading Networks with Integration Servers. I have several sites that are geographically located in different areas of North America.

I am hearing mixed thoughts on rather or not I should put an Integration Server at each location and have them all talk to one instance of Trading Networks or if I should use just one Integration Server and one instance of Trading Networks.

How are you using the combination of the two and what difficulties/issues have you seen in the way that you have it set up?

Having one IS per location will ensure that if one location goes down, other locations are not affected and are still up and running and processing msgs for that particular location.

From a TN perspective, you’re basically considering hub and spoke vs. federated (or peer-to-peer). Another approach is to centralize multiple IS/TN instances in a single data center.

Which approach fits best depends greatly on the integrations that will be hosted by the infrastructure.

I understand the fail-safe reasons for having one IS at each location. However would it not make more sense from a cost perspective if a few of the locations that are moderately low volume were put on one IS?

Cost-wise, the best approach is to have 2 IS/TN instances in a cluster (for failover purposes) in one location and have all locations use that one site.

However, other factors may push you to a distributed approach. Again, those factors are driven by what integrations will be hosted by the infrastructure, the service levels needed, anticipated volume, etc.

I’m not sure you’re going to get meaningful recommendations on which approach to use without providing more info about what the characteristics of the integrations.