Everything looks good on paper. When you're planning out a major new IT initiative, say a service-oriented architecture, you never create a space on the schematic which says "you will be plagued by consistent performance problems."

No one plans for that. Of course it could be argued that by not taking those potential issues into account during the conceptual phase you are, in a sense, planning for performance headaches. For instance, do you have a low latency strategy inside your SOA plan? When you're dealing with broadly distributed Web services and the verbosity of XML, high latency can rise up and bite you right on the butt. The best conceptual Web service on the planet can fall flat if it can't perform and scale.

Have you considered how multi-core processing can aid your SOA efforts? Hardware has become a big part of the SOA equation. The IT managers have a stake in this as well. They aren't going to get behind SOA projects if it means their lives become more difficult.